McCarthan v. Dir. of Goodwill Indus.-Suncoast, Inc.

Full title: Dan Carmichael MCCARTHAN, Petitioner-Appellant, v. DIRECTOR OF GOODWILL…

Court: United States Court of Appeals, Eleventh Circuit.

Date published: Mar 14, 2017

851 F.3d 1076 (11th Cir. 2017)

Opinion

No. 12-14989

03-14-2017

Dan Carmichael MCCARTHAN, Petitioner-Appellant, v. DIRECTOR OF GOODWILL INDUSTRIES-SUNCOAST, INC., Respondent-Appellee.

Sonya Rudenstine, Law Office of Sonya Rudenstine, GAINESVILLE, FL, for Petitioner-Appellant. Roberta Josephina Bodnar, U.S. Attorney's Office, ORLANDO, FL, Todd B. Grandy, Robert E. O'Neill, Sara C. Sweeney, TAMPA, FL, Lisa A. Hirsch, U.S. Attorney's Office, MIAMI, FL, for Respondent-Appellee. David Charles Frederick, Jeffrey A. Love, Kellogg, Hansen, Todd, Figel & Frederick, P.L.L.C., WASHINGTON, DC, Amicus Curiae for NATIONAL ASSOCIATION OF CRIMINAL DEFENSE LAWYERS. Taylor A.R. Meehan, Bartlit Beck Herman Palenchar & Scott, LLP, CHICAGO, IL, Court Appoint Amicus Curiae.

WILLIAM PRYOR, Circuit Judge

Sonya Rudenstine, Law Office of Sonya Rudenstine, GAINESVILLE, FL, for Petitioner-Appellant.

Roberta Josephina Bodnar, U.S. Attorney's Office, ORLANDO, FL, Todd B. Grandy, Robert E. O'Neill, Sara C. Sweeney, TAMPA, FL, Lisa A. Hirsch, U.S. Attorney's Office, MIAMI, FL, for Respondent-Appellee.

David Charles Frederick, Jeffrey A. Love, Kellogg, Hansen, Todd, Figel & Frederick, P.L.L.C., WASHINGTON, DC, Amicus Curiae for NATIONAL ASSOCIATION OF CRIMINAL DEFENSE LAWYERS.

Taylor A.R. Meehan, Bartlit Beck Herman Palenchar & Scott, LLP, CHICAGO, IL, Court Appoint Amicus Curiae.

Before ED CARNES, Chief Judge, TJOFLAT, HULL, MARCUS, WILSON, WILLIAM PRYOR, MARTIN, JORDAN, ROSENBAUM, JULIE CARNES, and JILL PRYOR, Circuit Judges.

WILLIAM PRYOR, Circuit Judge:

This appeal requires us to decide whether a change in caselaw entitles a federal prisoner to an additional round of collateral review of his sentence. Congress gives a federal prisoner like Dan McCarthan one opportunity to move to vacate his sentence unless that remedy is "inadequate or ineffective to test the legality of his detention." 28 U.S.C. § 2255(e). When McCarthan pleaded guilty to being a felon in possession of a firearm, 18 U.S.C. § 922(g), he understood that the district court would enhance his sentence under the Armed Career Criminal Act, id. § 924(e). He did not appeal that sentence. When McCarthan later moved to vacate his sentence, he again said nothing about the enhancement. After foregoing those opportunities to complain about the enhancement of his sentence, McCarthan petitioned for a writ of habeas corpus. McCarthan argues that his earlier motion to vacate was inadequate to test his objection to his sentence enhancement because our caselaw about the Armed Career Criminal Act has changed. But because the motion to vacate gave McCarthan an opportunity to challenge his sentence enhancement, his remedy was not inadequate or ineffective to test the legality of his sentence, regardless of any later change in caselaw.

For eighteen years, our Court has maintained that a change in caselaw may trigger an additional round of collateral review, see Wofford v. Scott , 177 F.3d 1236 (11th Cir. 1999), but our precedents have ignored the text of the statute. As we struggled to apply our precedents, we employed a five-factor test and granted relief only twice. See Mackey v. Warden, FCC Coleman-Medium , 739 F.3d 657 (11th Cir. 2014) ; Bryant v. Warden, FCC Coleman-Medium , 738 F.3d 1253 (11th Cir. 2013). Because our precedents have failed to adhere to the text of section 2255(e), have not incurred significant reliance interests, and have proved unworkable, today we overrule them. We join the Tenth Circuit in applying the law as Congress wrote it, see Prost v. Anderson , 636 F.3d 578 (10th Cir. 2011) (Gorsuch, J.), and hold that a change in caselaw does not make a motion to vacate a prisoner's sentence "inadequate or ineffective to test the legality of his detention," 28 U.S.C. § 2255(e). We affirm the dismissal of McCarthan's petition for a writ of habeas corpus.

I. BACKGROUND

In 2003, Dan McCarthan pleaded guilty to being a felon in possession of a firearm, 18 U.S.C. § 922(g), the maximum sentence for which is ten years imprisonment, id. § 924(a)(2). The district court enhanced McCarthan's sentence under the Armed Career Criminal Act, id. § 924(e), on the ground that he had five prior convictions for a "serious drug offense" or a "violent felony," id. § 924(e)(1), including one for escape. United States v. McCarthan , No. 8:02-cr-137 (M.D. Fla. June 4, 2003). McCarthan received a sentence of 211 months. Id. He did not appeal. Id.

McCarthan later moved to vacate his sentence, 28 U.S.C. § 2255. He alleged that he had received ineffective assistance of counsel, but he did not challenge the enhancement of his sentence. The district court denied the motion to vacate on the merits. McCarthan v. United States , No. 8:04-cv-1288 (M.D. Fla. Sept. 30, 2004). Both the district court and this Court denied his request for a certificate of appealability. See id.

In 2009, the Supreme Court ruled that some forms of the crime of escape do not qualify as a "violent felony" under the Armed Career Criminal Act. Chambers v. United States , 555 U.S. 122, 130, 129 S.Ct. 687, 172 L.Ed.2d 484 (2009). Chambers overturned our circuit precedent, United States v. Gay , 251 F.3d 950 (11th Cir. 2001), that even "walkaway" escape qualified as a violent felony. Id. at 954–55. Because Chambers involved statutory interpretation, McCarthan could not bring a second motion to vacate under section 2255(h). Instead, he filed a petition for a writ of habeas corpus, 28 U.S.C. § 2241. Both the district court and the panel applied a test we first enunciated in Wofford that would allow a federal prisoner to petition for a writ of habeas corpus if a later decision of the Supreme Court abrogates circuit precedent that had foreclosed the prisoner's argument when he first moved to vacate his sentence.

The district court could have exercised jurisdiction over McCarthan's petition only if it fell within the saving clause of section 2255(e). McCarthan argued that Chambers "ma[de] [him] actually innocent" of the sentencing enhancement and made him eligible for relief under the saving clause. The district court dismissed the petition because McCarthan's other convictions ensured that his sentence did not exceed the statutory maximum. McCarthan v. Warden, FCC Coleman-Medium , 5:09-cv-110 (M.D. Fla. Jan. 11, 2012).

We affirmed the dismissal of McCarthan's petition. McCarthan v. Warden, FCI Estill , 811 F.3d 1237, 1242 (11th Cir. 2016), reh'g en banc granted, op. vacated , No. 12-14989, 2016 WL 3878151 (11th Cir. May 24, 2016). The panel opinion explained that McCarthan's petition did not satisfy the requirements of the Wofford test because he had at least three other convictions that triggered his enhanced sentence. Id. at 1256–57. But the panel disagreed about how to apply the Wofford test. Compare id. at 1246–47, with id. at 1257–59 (Proctor, J., concurring).

McCarthan filed a petition for rehearing en banc, and we granted it. We instructed the parties to brief three issues: (1) do our precedents erroneously interpret the saving clause, 28 U.S.C. § 2255(e) ; (2) what is the correct interpretation of the saving clause; and (3) applying the correct standard, is McCarthan entitled to petition for a writ of habeas corpus? Because both McCarthan and the Warden argued that the Wofford test or some version of it is correct, we appointed Taylor Meehan as amicus curiae to argue that our precedents erroneously interpreted the saving clause. We thank Ms. Meehan for her superb brief and oral argument in keeping with the highest tradition of the legal profession.

On October 17, 2016, we granted McCarthan's unopposed motion to substitute the Director of Goodwill Industries-Suncoast, Inc. as the Respondent-Appellee. McCarthan was transferred from FCI Estill to the custody of the Director of Goodwill Industries-Suncoast, Inc., a Bureau of Prisons Residential Reentry Center (more commonly known as a halfway house). McCarthan is still "in custody," for purposes of our jurisdiction. 28 U.S.C. § 2255(a).

II. STANDARD OF REVIEW

Whether a prisoner may bring a petition for a writ of habeas corpus under the saving clause of section 2255(e) is a question of law we review de novo . Williams v. Warden, Federal Bureau of Prisons , 713 F.3d 1332, 1337 (11th Cir. 2013). The petitioner bears the burden of establishing that the remedy by motion was "inadequate or ineffective to test the legality of his detention." Turner v.Warden Coleman FCI (Medium) , 709 F.3d 1328, 1333 (11th Cir. 2013) (quoting 28 U.S.C. § 2255(e) ), abrogated on other grounds by Johnson v. United States , ––– U.S. ––––, 135 S.Ct. 2551, 192 L.Ed.2d 569 (2015).

III. DISCUSSION

Since 1948, Congress has required that a federal prisoner file a motion to vacate, 28 U.S.C. § 2255, instead of a petition for a writ of habeas corpus, id. § 2241, to collaterally attack the legality of his sentence. See Pub. L. No. 80-773, 62 Stat. 869, 967–68. A motion to vacate allows a prisoner to contest his sentence "upon the ground that the sentence was imposed in violation of the Constitution or laws of the United States, or that the court was without jurisdiction to impose such sentence, or that the sentence was in excess of the maximum authorized by law, or is otherwise subject to collateral attack." 28 U.S.C. § 2255(a). Section 2255(e) makes clear that a motion to vacate is the exclusive mechanism for a federal prisoner to seek collateral relief unless he can satisfy the "saving clause" at the end of that subsection:

An application for a writ of habeas corpus in behalf of a prisoner who is authorized to apply for relief by motion pursuant to this section, shall not be entertained if it appears that the applicant has failed to apply for relief, by motion, to the court which sentenced him, or that such court has denied him relief, unless it also appears that the remedy by motion is inadequate or ineffective to test the legality of his detention.

Id. § 2255(e) (emphasis added). "[S ]aving [, not savings,] is the precise word" for "a statutory provision exempting from coverage something that would otherwise be included," Bryan A. Garner, Garner's Dictionary of Legal Usage 797 (3d ed. 2011); it has nothing to do with saving a statute from unconstitutionality, see, e.g. , 28 U.S.C. § 1333(1) ("saving to suitors in all cases all other remedies to which they are otherwise entitled").

To determine whether a change in caselaw can satisfy the saving clause of section 2255(e), we consider three matters. First, we explain how we (and other circuits) have interpreted the saving clause. Second, we explain why our precedents fail to adhere to the text of the saving clause. Third, in the light of the incongruity of the text and our precedents, we explain our decision to overrule our precedents.

A. Our Precedents About the Saving Clause

Congress enacted section 2255 to address the "serious administrative problems" caused by the requirement that habeas petitions be brought in the district of incarceration, often far from where relevant records and witnesses were located. United States v. Hayman , 342 U.S. 205, 210–19, 72 S.Ct. 263, 96 L.Ed. 232 (1952). The motion to vacate "afford[ed] the same rights in another and more convenient forum," namely the district where the prisoner was sentenced. Id. at 219, 72 S.Ct. 263. In 1996, Congress reformed the system of collateral review when it passed the Antiterrorism and Effective Death Penalty Act. See Pub. L. No. 104-132, 110 Stat. 1214. The Act made several changes to section 2255, including the addition of a bar on second or successive motions, 28 U.S.C. § 2255(h), and a statute of limitations, id. § 2255(f). See 110 Stat. at 1220. But the Act did not alter the saving clause. See id.

This Circuit first considered the meaning of the saving clause eighteen years ago in Wofford . Charlie Wofford, a federal prisoner, pleaded guilty to being a felon in possession of a firearm and conspiracy to possess with intent to distribute cocaine. Wofford , 177 F.3d at 1237. The district court and this Court denied his first motion to vacate. Id. After several failed attempts to file successive motions to vacate, Wofford petitioned for a writ of habeas corpus under section 2241 and argued that his illegal sentence created manifest injustice. Id. at 1238. Wofford argued that because the bar on second and successive motions prevented the court from reaching the merits of his new claims, he satisfied the saving clause. Id.

We denied Wofford relief, but our analysis paid scant attention to the text of the saving clause. We began with the opinion of the Supreme Court in Hayman , but concluded that it was "not very helpful" with respect to the "saving [ ] clause language." Id. at 1239. We then discussed the legislative history. Id. at 1239–41. Early versions of the saving clause focused on practicable problems, but we "found nothing in the legislative history explaining why the relevant language was changed or what the new language means." Id. at 1241. Unsurprisingly, snippets from the legislative history cut both ways—that the new language did not make any substantive changes and that the new language was broader than the old language—but we decided "the better view is that the saving[ ] clause is concerned with more than the practical difficulties." Id. We then canvassed the decisions of our sister circuits. Id. at 1242–45. After we concluded that the approach of the Seventh Circuit was "better reasoned than those of the other circuits, and its rule has the advantage of being specific," we applied a test that turned on an intervening change in circuit precedent. Id. at 1244. We stated that "the only sentencing claims that may conceivably be covered by the saving[ ] clause are those based upon a retroactively applicable Supreme Court decision overturning circuit precedent." Id. at 1245. But because Wofford's petition did not rest upon a change in caselaw, we denied him relief. Id.

In Gilbert , sitting en banc, we clarified that, under the Wofford test, the saving clause does not apply to errors that do not cause a sentence to exceed the statutory maximum. Gilbert v. United States , 640 F.3d 1293, 1323 (11th Cir. 2011) (en banc). Ezell Gilbert pleaded guilty to possession with intent to distribute of more than 50 grams of crack cocaine and more than 100 grams of marijuana. Id. at 1298. Under the then-mandatory sentencing guidelines, the district court applied the career offender enhancement and sentenced Gilbert to 292 months imprisonment. Id. at 1300. The statutory maximum was life imprisonment. Id. at 1299. Years after we denied relief in his direct appeal and denied him a certificate of appealability about the denial of his first motion to vacate, Gilbert invoked the saving clause and petitioned for a writ of habeas corpus. He argued that the district court should not have applied the career offender guideline. Id. at 1301–02. On rehearing en banc, we explained that the Wofford test was "only dicta" but, in any event, could not help Gilbert. Id. at 1319–20. Because Gilbert challenged only the use of the guidelines in determining his sentence and a prisoner cannot be actually innocent of a sentence within the statutory range, Wofford did not apply. Id. at 1320 ("Gilbert's position turns on treating sentences as convictions, and an argument that depends on calling a duck a donkey is not much of an argument."). Because Gilbert's sentence did not exceed the statutory maximum, we denied relief. Id. at 1322–24.

A few years later, Williams revisited the Wofford test to address an alleged error that caused the sentence to exceed the statutory maximum. Williams , 713 F.3d at 1334. Albert Williams was tried and convicted of being a felon in possession of a firearm and was sentenced as a career offender under the Armed Career Criminal Act. Id. at 1335. Williams did not object to the enhancement during sentencing or on direct appeal, and we affirmed his conviction. Id. After Williams filed several meritless collateral attacks, the Supreme Court decided Begay v. United States , 553 U.S. 137, 128 S.Ct. 1581, 170 L.Ed.2d 490 (2008), which narrowed the definition of "violent felony" in the Armed Career Criminal Act. Id. at 148, 128 S.Ct. 1581. Williams argued that under section 2255(e) the district court could hear his petition for a writ of habeas corpus and decide his claim that his sentence now exceeded the statutory maximum because some of his underlying convictions no longer qualified as violent felonies. Williams , 713 F.3d at 1336. We reiterated that "the statute says precious little about what it means for the original motion to have been 'inadequate' or 'ineffective.' " Id. at 1341. Applying the Wofford test, we determined that circuit precedent would not have "squarely resolved" Williams's claim unless there was "adverse precedent ... that would have made us unwilling to listen." Id. at 1343, 1347. But there was "no Circuit precedent on the books during Williams's collateral attack" that foreclosed his claim, so we denied him relief. Id. at 1345, 1349.

In Bryant , we again applied the Wofford test and granted a prisoner relief under the saving clause for the first time. Bryant , 738 F.3d at 1274. We distilled a five-part test from our precedents. That is, a federal prisoner may file a petition for a writ of habeas corpus if (1) binding precedent foreclosed a claim at the time of his first motion to vacate; (2) the Supreme Court overturned our binding precedent that foreclosed the claim; (3) the new decision of the Supreme Court applies retroactively on collateral review; (4) as a result of this retroactive decision, the prisoner's sentence is now contrary to the law; and (5) this kind of claim can be brought under the saving clause. Id. Bryant pleaded guilty to being a felon in possession of a firearm. Id. at 1258. He had three prior felony convictions, including one for carrying a concealed firearm, and the district court imposed a sentencing enhancement under the Armed Career Criminal Act. Id. at 1258–60. Bryant's first motion to vacate did not challenge his conviction for carrying a concealed firearm. Id. at 1260. After Begay , the district court denied leave to file a successive motion, and Bryant instead petitioned for a writ of habeas corpus. Id.

Bryant satisfied each part of the Wofford test. Our precedent in United States v. Hall , 77 F.3d 398 (11th Cir. 1996), held that a concealed-firearm offense was a violent felony, id. at 401–02, which foreclosed Bryant's argument when he filed his first motion to vacate. Bryant , 738 F.3d at 1274. The later decision of the Supreme Court in Begay "busted" that precedent. Id. at 1275. And we held that Begay announced a substantive new rule that applied retroactively. Id. at 1276–77. Bryant's sentence was 235 months imprisonment, which exceeded the ten year statutory maximum for his crime without the enhancement. Id. at 1279. And we held that the saving clause reaches more than claims of actual innocence; it extends also to errors that cause a sentence to exceed the statutory maximum. Id. at 1281–84.

In a similar appeal, we also granted Brian Mackey relief. Mackey , 739 F.3d at 663. He argued that as a result of Begay , his convictions for carrying a concealed firearm no longer supported his sentence as an armed career criminal. Id. at 660. Under the Wofford test as explicated in Bryant , we again granted relief. Id. at 658, 663.

Since then, additional wrinkles have arisen. In Samak , a federal inmate imprisoned in our Circuit, but sentenced in another, the Fifth, petitioned for a writ of habeas corpus. Samak v. Warden, FCC Coleman-Medium , 766 F.3d 1271, 1275 n.3 (11th Cir. 2014). The Wofford test required us to review Fifth Circuit precedent and determine whether the law of that other circuit foreclosed Jamal Samak's petition. Id. at 1275. We denied relief because the relevant Fifth Circuit precedent actually supported Samak's claim at the time of his first motion to vacate. Id. But a separate concurring opinion called for a reconsideration of our precedent in Bryant and the adoption of an interpretation rooted in the plain text of the saving clause. Id. at 1275–76 (W. Pryor, J., concurring). And in Cortes-Morales , a federal prisoner argued that the saving clause should extend beyond changes in caselaw to retroactive legislation that amended the New York sentencing statutes. Cortes-Morales v. Hastings , 827 F.3d 1009, 1015 (11th Cir. 2016). We held that Jorge Cortes-Morales was not eligible for resentencing under the revised New York laws and avoided the question whether or not the saving clause could be extended to retroactive amendments to state legislation. Id. at 1016. But a separate concurring opinion reiterated that because Bryant is a "monster of our creation, untethered to the text" there is "no principled basis for determining its ultimate reach." Id. (W. Pryor, J., concurring).

Several other circuits have divined similarly atextual tests for satisfying the saving clause. In Davenport , which we cited in Wofford , the Seventh Circuit engaged in a pragmatic analysis that adequate "should mean" that "a prisoner [has] a reasonable opportunity to obtain a reliable judicial determination of the fundamental legality of his conviction and sentence." In re Davenport , 147 F.3d 605, 609 (7th Cir. 1998). And the Fourth, Fifth, and Sixth Circuits have required proof of actual innocence of a charged offense, in addition to other factors, to obtain relief under the saving clause. See, e.g. , Wooten v. Cauley , 677 F.3d 303, 307–08 (6th Cir. 2012) ; Reyes–Requena v. United States , 243 F.3d 893, 904 (5th Cir. 2001) ; In re Jones , 226 F.3d 328, 333–34 (4th Cir. 2000). The Second Circuit holds that "inadequate or ineffective" means "the set of cases in which the petitioner cannot, for whatever reason, utilize § 2255, and in which the failure to allow for collateral review would raise serious constitutional questions." Triestman v. United States , 124 F.3d 361, 377 (2d Cir. 1997). The Third Circuit focuses on when the second or successive limitations would cause a "complete miscarriage of justice." In re Dorsainvil , 119 F.3d 245, 251 (3d Cir. 1997). And in the Eighth and Ninth Circuits, a prisoner must not have had an unobstructed procedural shot at presenting that claim, defined to include changes in law. See Harrison v. Ollison , 519 F.3d 952, 959–60 (9th Cir. 2008) ; Abdullah v. Hedrick , 392 F.3d 957, 963 (8th Cir. 2004). Judge Martin's dissent places great weight on the majority of circuits having arrived at the same result, regardless of their reasoning, Martin Dissent at 1113–14, 1118 n.7, but our inquiry must begin with the text.

Only the Tenth Circuit has adhered to—or even seriously considered—the text of the saving clause. In Prost , the Tenth Circuit held that "the plain language of § 2255 means what it says and says what it means: a prisoner can proceed to § 2241 only if his initial § 2255 motion was itself inadequate or ineffective to the task of providing the petitioner with a chance to test his sentence or conviction." Prost , 636 F.3d at 587. The intervening change in caselaw does not mean that the "process was ineffective or inadequate to test his argument." Id. at 580. And then-Chief Judge Frank Easterbrook reached the same conclusion contrary to the circuit precedent that binds his court: "A motion under § 2255 could reasonably be thought 'inadequate or ineffective to test the legality of [the prisoner's] detention' if a class of argument were categorically excluded, but when an argument is permissible but fails on the merits there is no problem with the adequacy of § 2255." Brown v. Caraway , 719 F.3d 583, 597 (7th Cir. 2013) (Easterbrook, C.J., concerning the circulation under Circuit Rule 40(e)).

In Bryant , we briefly considered this textual interpretation of the saving clause and dismissed it as "in tension with this Court's precedent." 738 F.3d at 1287. But as the Tenth Circuit correctly explained, our precedent in Wofford did not address the "textual and structural clues" that support the contrary reasoning in Prost, 636 F.3d at 593. With the benefit of our experience, we now take this opportunity to reconsider our interpretation of the saving clause.

B. The Text of the Saving Clause

The saving clause provides a federal prisoner relief only when his "remedy by motion is inadequate or ineffective to test the legality of his detention." 28 U.S.C. § 2255(e). When we read this text, several terms offer important clues about its meaning: "remedy," "to test," "inadequate or ineffective," and "detention." Careful review of these terms and the whole text makes clear that a change in caselaw does not trigger relief under the saving clause. Whether circuit precedent "was once adverse to a prisoner has nothing to do with whether his motion to vacate his sentence is 'inadequate or ineffective to test the legality of his detention.' " Samak , 766 F.3d at 1276 (W. Pryor, J., concurring).

McCarthan's claim that his sentence exceeds the statutory maximum is exactly the kind of claim that a motion to vacate is designed to "remedy," notwithstanding adverse precedent. "Remedy" as used in the saving clause does not promise "relief." A "remedy" is "[t]he means by which a right is enforced or the violation of a right is prevented, redressed, or compensated." Remedy , Black's Law Dictionary 1526 (3d ed. 1933). "Relief" is "the assistance, redress, or benefit which a complainant seeks at the hands of the court." Relief , Black's Law Dictionary 1523 (3d ed. 1933). The "means" are not inadequate when circuit precedent forecloses relief on a claim. The remedy of a motion to vacate permitted McCarthan to bring his claim and seek en banc or Supreme Court review to change the substantive rule of law. That a court might reject a prisoner's argument does not render his "remedy by motion" an inadequate "means by which" to challenge the legality of his sentence. A procedural rule that might prevent success on a particular motion does not render the remedy an inadequate "means" so long as it is capable of "enforc [ing]" or "redress[ing]" the right. The motion to vacate is an adequate remedy for McCarthan because if he succeeds, the court must "vacate and set the judgment aside" and either release or retry him. 28 U.S.C. § 2255(b).

The distinction between remedy and relief is reflected throughout our system of habeas corpus. For example, a procedural bar might prevent relief, but that bar does not render the motion itself an ineffective or inadequate remedy. See , e.g. , Jiminian v. Nash , 245 F.3d 144, 147–48 (2d Cir. 2001) (Sotomayor, J.). The prisoner may still bring the claim. Likewise, a state prisoner must "exhaust[ ] the remedies available in the courts of the State" before petitioning for a federal writ of habeas corpus. 28 U.S.C. § 2254(b)(1)(A) (emphasis added). In this context, remedy must refer to the available process—not substantive relief—because a prisoner who received relief in state court would have no reason to file a habeas petition. That McCarthan's argument was foreclosed by precedent (as opposed to being wrong, untimely, procedurally barred, or unexhausted) is irrelevant. The motion to vacate provided an adequate remedy to challenge the legality of his sentence.

McCarthan also could have "tested" the legality of his detention in his first motion to vacate. That is, he could have made the argument that his prior convictions did not qualify him for an enhanced sentence under the statute. "To test" the legality of his detention and satisfy the saving clause, a prisoner is not required "to win" his release. "To test" means "to try." Test , 11 Oxford English Dictionary 220 (1st ed. 1933). To try a claim, a "petitioner [must have] an opportunity to bring his argument," Prost , 636 F.3d at 584. The opportunity to test or try a claim, however, neither guarantees any relief nor requires any particular probability of success; it guarantees access to a procedure. Id. ; see also Taylor v. Gilkey , 314 F.3d 832, 835–36 (7th Cir. 2002) ("[To test] implies a focus on procedures rather than outcomes. Judges sometimes err, but this does not show that the procedures are inadequate; it shows only that people are fallible."). To determine whether a prisoner satisfies the saving clause, we ask only whether the motion to vacate is an adequate procedure to test the prisoner's claim. And to answer this question, we ask whether the prisoner would have been permitted to bring that claim in a motion to vacate. In other words, a prisoner has a meaningful opportunity to test his claim whenever section 2255 can provide him a remedy.

Despite circuit precedent, McCarthan could have tested the legality of his detention by requesting that we reconsider our precedent en banc or by petitioning the Supreme Court for a writ of certiorari. The panel opinion stated that the purpose of the Wofford test is "to prevent us from entertaining § 2241 petitions by federal prisoners who could have at least theoretically successfully challenged an ACCA enhancement in an earlier proceeding," McCarthan , 811 F.3d at 1245, and Judge Rosenbaum's dissent argues that "a prisoner must have a 'meaningful opportunity' to present his claim," Rosenbaum Dissent at 1135–36. But if McCarthan had raised his claim earlier, perhaps he could have been the successful litigant that Deondery Chambers or Larry Begay later came to be. For example, Chambers raised the same claim McCarthan does, namely that his conviction for escape was not a violent felony under the Armed Career Criminal Act. Chambers , 555 U.S. at 123, 129 S.Ct. 687. And he too faced binding circuit precedent that foreclosed this claim. See United States v. Chambers , 473 F.3d 724, 725–26 (7th Cir. 2007), rev'd , 555 U.S. 122, 129 S.Ct. 687, 172 L.Ed.2d 484 (2009). But he nevertheless presented his claim and won relief in the Supreme Court. Chambers , 555 U.S. at 126–27, 129 S.Ct. 687. Similarly, in the context of procedural default, we do not excuse a defendant's failure to raise a claim even if the claim was "unacceptable to that particular court at that particular time." Engle v. Isaac , 456 U.S. 107, 132 n.35, 102 S.Ct. 1558, 71 L.Ed.2d 783 (1982) (citation omitted); Moore v. Zant , 885 F.2d 1497, 1507–08 (11th Cir. 1989) (" Engle ... indicated that petitioners might have a duty to anticipate changes in the law at the threat of having later claims based on those changes barred by principles of procedural default."). It is unclear why the chance to have precedent overruled en banc or by the Supreme Court would not qualify as a theoretically successful challenge or meaningful opportunity. McCarthan, like Chambers, had a meaningful opportunity to present his claim and test the legality of his sentence before the court of appeals and before the Supreme Court. A test often failed can nevertheless be an adequate test.

Adverse circuit precedent also did not make McCarthan's first motion to vacate his sentence "inadequate or ineffective" to challenge his sentence. "Inadequate or ineffective" instead connotes that the saving clause permits a prisoner to bring a claim in a petition for habeas corpus that could not have been raised in his initial motion to vacate. The term "inadequate," as defined in the phrase "inadequate remedy at law," means "unfitted or not adapted to the end in view." Inadequate Remedy at Law , Black's Law Dictionary 940 (3d ed. 1933); see also Jordan Concurring at 1105 (providing a definition of "inadequate" as "lacking in effectiveness"). And "ineffective" means "[o]f such a nature as not to produce ... the intended [ ] effect." Ineffective , 5 Oxford English Dictionary 239 (1st ed. 1933). That a particular argument is doomed under circuit precedent says nothing about the nature of the motion to vacate. The motion to vacate is still "adapted to the end" of testing the claim regardless of the claim's success on the merits.

The word "or" in "inadequate or ineffective" does not overpower the ordinary meaning of the words, which have similar definitions. We are hard pressed to imagine a remedy that is "lacking in effectiveness" but not "ineffective," or "of such a nature as not to produce the intended effect" but not "inadequate." Although the disjunctive "or" may suggest separate meanings for the two terms, Jordan Concurring at 1104–05; Rosenbaum Dissent at 1130, it does not require mutual exclusivity. The word "or" commonly introduces a synonym or "definitional equivalent." See Antonin Scalia & Bryan A. Garner, Reading Law: An Interpretation of Legal Texts 122 (2012). That construction may be an example of the "ill-conceived but lamentably common belt-and-suspenders approach" to legal writing, id. at 176–77, but it is the better reading of the text when the terms share the same ordinary meaning. Judge Rosenbaum's dissent disagrees because the phrase "or ineffective" is not set off by commas, Rosenbaum Dissent at 1130, but commas are not necessary. See, e.g. , Scalia & Garner, supra , at 122 ("The award of exemplary or punitive damages is the exception, not the rule."); Fed. R. Evid. 407 ("When measures are taken that would have made an earlier injury or harm less likely to occur ..."). That the definitions overlap does not require that we ignore the ordinary meaning of the text, and it does not support the dissent's conclusion that "ineffective" must mean "constitutionally deficient." Rosenbaum Dissent at 1130–31.

A motion to vacate is not often an inadequate or ineffective remedy. But a motion to vacate could be "inadequate or ineffective to test" a prisoner's claim about the execution of his sentence because that claim is not cognizable under section 2255(a). See, e.g. , Hajduk v. United States , 764 F.2d 795, 796 (11th Cir. 1985). The motion to vacate is "of such a nature" that it will "not ... produce ... the intended [ ] effect," Ineffective , 5 Oxford English Dictionary 239 (1st ed. 1933), because the prisoner does not challenge his sentence and the appropriate remedy is not vacatur. Or, if the sentencing court no longer exists, the remedy by motion could be "inadequate or ineffective to test" the prisoner's claim because the motion may be brought only in that venue. But when a prisoner's argument about the legality of his sentence conflicts with circuit precedent, a motion to vacate is neither inadequate nor ineffective to test his argument.

The word "ineffective" also carries this meaning elsewhere in the statute: a state prisoner may avoid the exhaustion requirements if "circumstances exist that render such process ineffective to protect the rights of the applicant." 28 U.S.C. § 2254(b)(1)(B)(ii). The Supreme Court stated that this exception applies only if "there is no opportunity to obtain redress in state court or if the corrective process is so clearly deficient as to render futile any effort to obtain relief ." Duckworth v. Serrano , 454 U.S. 1, 3, 102 S.Ct. 18, 70 L.Ed.2d 1 (1981) (emphasis added). Because there was no claim that the "postconviction procedures [were] inadequate to adjudicate" the claim, the prisoner did not qualify for the exception. Id. at 4, 102 S.Ct. 18. So too here. The remedy by motion is not ineffective unless the procedure it provides is incapable of adjudicating the claim. We cannot "engraft[ ] an exception onto the habeas statute not envisioned by Congress [and] inconsistent with the clear mandate of the Act." Id. at 5, 102 S.Ct. 18.

In other areas of the law, adequacy and effectiveness focus on process and do not require any likelihood of success on the merits. For example, a litigant with an "adequate" remedy at law cannot seek equitable relief, even if his legal claim has little chance of success. Samak , 766 F.3d at 1285 (W. Pryor, J., concurring). Similarly, in the context of the Sixth Amendment, defense counsel is not "ineffective" even if his arguments are "doomed." Brown , 719 F.3d at 597 (Easterbrook, C.J., concerning the circulation under Circuit Rule 40(e)). Judge Rosenbaum's dissent takes this analogy too far when it asserts that because the ineffective assistance of counsel creates a constitutional deficiency under the Sixth Amendment, the term "ineffective" means "constitutionally deficient." Rosenbaum Dissent at 1132.

When circuit precedent forecloses a prisoner's claim, "it may very well mean circuit law is inadequate or deficient. But that does not mean the § 2255 remedial vehicle is inadequate or ineffective to the task of testing the argument." Prost , 636 F.3d at 590. A prisoner has an adequate procedure to raise any claim attacking his sentence, even if that claim is foreclosed by circuit precedent. Our precedent may later prove to be "right or wrong as a matter of substantive law, but the saving[ ] clause is satisfied so long as the petitioner had an opportunity to bring and test his claim." Id. at 585. When a prisoner's motion attacks his sentence based on a cognizable claim that can be brought in the correct venue, the remedy by motion is adequate and effective to test his claim.

The term "detention" in the saving clause carries a broader meaning than the term "sentence" that appears elsewhere in the statute. Section 2255(a) allows a prisoner to challenge only his "sentence." 28 U.S.C. § 2255(a). But the saving clause preserves challenges to a prisoner's "detention" that would otherwise go unremedied. Id. § 2255(e). When Congress uses "different language in similar sections," we should give those words different meanings. See Iraola & CIA, S.A. v. Kimberly–Clark Corp. , 232 F.3d 854, 859 (11th Cir. 2000) ; see also Scalia & Garner, supra , at 170. When Congress enacted section 2255, the word "detention" meant "[k]eeping in custody or confinement," Detention , 3 Oxford English Dictionary 266 (1st ed. 1933), or "[t]he act of keeping back or withholding, either accidentally or by design, a person or thing," Detention , Black's Law Dictionary 569 (3d ed. 1933). Because someone can be "[kept] in custody" without a criminal sentence, or "with[eld]" contrary to the terms of the sentence, it is clear that the meaning of "detention" covers circumstances of confinement other than those attributable to the sentence.

When a prisoner attacks aspects of his detention in ways that do not challenge the validity of his sentence, then the saving clause may provide him access to a different remedy. For example, a prisoner may concede the validity of his sentence but raise claims about the execution of his sentence—that is, "about his good-time credits or the revocation of his parole, which involve the 'act of keeping back or withholding' the prisoner." Samak , 766 F.3d at 1280 (W. Pryor, J., concurring).

This reading of the text comports with the traditional distinction between a motion to vacate and a petition for a writ of habeas corpus. A motion to vacate covers only challenges to the validity of a sentence, but the saving clause and a petition for a writ of habeas corpus cover challenges to the execution of a sentence. Cf. Antonelli v. Warden, U.S.P. Atlanta , 542 F.3d 1348, 1351 n.1 (11th Cir. 2008) ("It is well-settled that a § 2255 motion to vacate is a separate and distinct remedy from habeas corpus proper.... A prisoner in custody pursuant to a federal court judgment may proceed under § 2241 only when he raises claims outside the scope of § 2255(a), that is, claims concerning execution of his sentence." (internal citations omitted)); United States v. Flores , 616 F.2d 840, 842 (5th Cir. 1980) ("[The prisoner's] appropriate remedy is under § 2255, not 28 U.S.C. § 2241, since the alleged errors occurred at or prior to sentencing."). Because Congress used "sentence" in one part of the statute and "detention" in another, we should interpret the statute to preserve the traditional distinction between those terms and the procedures by which they are challenged.

McCarthan's petition does not fall within the text of the saving clause. Nothing in the text suggests that Congress gave special status to claims foreclosed by binding circuit precedent, as opposed to claims that are procedurally defaulted or substantively wrong. See Samak , 766 F.3d at 1295 (W. Pryor, J., concurring) (" Bryant does not even attempt to offer a plausible interpretation of the text of the saving[ ] clause."). Neither McCarthan's failure to bring this claim earlier nor his odds of success on the merits are relevant to the saving clause inquiry. Because McCarthan filed a traditional claim attacking his sentence that he could have brought in a motion to vacate, the remedy by motion is adequate and effective to test the legality of his detention.

The whole text of section 2255 confirms our reading of the saving clause. "[T]here can be no justification for needlessly rendering provisions in conflict if they can be interpreted harmoniously." Scalia & Garner, supra , at 180. Allowing a prisoner with a claim that is cognizable in a motion to vacate to access the saving clause nullifies the procedural hurdles of section 2255 and undermines the venue provisions.

If the saving clause "guarantee[d] multiple opportunities to test a conviction or sentence," then the bar against second and successive motions under section 2255(h) would become a nullity. Prost , 636 F.3d at 585. Only prisoners who satisfy the exceptions of section 2255(h) may collaterally attack their sentences more than once. Section 2255(h)"speaks directly" to the question of "[h]ow often to rerun a search for error." Taylor , 314 F.3d at 835. Judge Martin's dissent argues that our interpretation "has made a rule" that "insulate[s] [our] mistakes from ... review," Martin Dissent at 1119, but, as always, every error we make in affirming the denial of a motion to vacate is subject to review on petition for certiorari. And we did not make the rule that bars consideration of second or successive motions. Congress did. The legislative branch defined both the appropriate sentence for McCarthan's crime and the rules for challenging the legality of that sentence. Congress recognized that courts would make mistakes, but provided for successive motions only in specific circumstances. The statute limits each prisoner to a "single collateral attack, unless the conditions of [2255(h) ] have been met." Taylor , 314 F.3d at 835. McCarthan neither alleges that "newly discovered evidence" establishes his innocence nor that "a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable" warrants relief. 28 U.S.C. § 2255(h). He cannot bring a second collateral attack.

The saving clause does not create a third exception. "Where Congress explicitly enumerates certain exceptions to a general prohibition, additional exceptions are not to be implied." TRW Inc. v. Andrews , 534 U.S. 19, 28, 122 S.Ct. 441, 151 L.Ed.2d 339 (2001) (quoting Andrus v. Glover Constr. Co. , 446 U.S. 608, 616–17, 100 S.Ct. 1905, 64 L.Ed.2d 548 (1980) ). The specific language of section 2255(h), enacted nearly 50 years after the saving clause, limits the reach of the saving clause. See Gilbert , 640 F.3d at 1308 ("An ambiguous or general statutory provision enacted at an earlier time must yield to a specific and clear provision enacted at a later time."). If Congress wanted an exception for all intervening changes in law, it could have said so. Elsewhere in the statute, Congress refers to any "right" that is new and retroactively applicable. 28 U.S.C. § 2255(f)(3). But section 2255(h) speaks only to "a new rule of constitutional law." Id. § 2255(h)(2). This material variation in terms suggests a variation in meaning. See Scalia & Garner, supra , at 170. Judge Martin's dissent argues that Congress's failure to repeal the saving clause permits courts to create a third exception for new rules of statutory interpretation that arise after a prisoner has used his first motion to vacate. Martin Dissent at 1117. But to read the bar on successive motions (or other procedural bars to relief) to trigger the saving clause makes the statute self-defeating. See, e.g. , Brown , 719 F.3d at 599 (Easterbrook, C.J., concerning the circulation under Circuit Rule 40(e)). And we are not persuaded that relying on equity to limit the third exception to claims of actual innocence, Jordan Concurring at 1105–08, does any less violence to the statutory text that creates only two exceptions. Congress did not create any exception to section 2255(h) for non-constitutional changes in law, so we may not craft one.

Section 2255 includes other procedural hurdles that the Wofford test fails to respect. For example, the Wofford test runs roughshod over the statute of limitations, 28 U.S.C. § 2255(f). A federal prisoner has one year to move to vacate his sentence under section 2255. But when a prisoner uses the saving clause to bring a claim that is cognizable in a motion to vacate, he bypasses his statute of limitations and gains limitless time to press claims that prisoners who meet the requirements of section 2255 do not receive.

The motion to vacate was intended to be a substitute remedy for the writ of habeas corpus, see Hill v. United States , 368 U.S. 424, 427, 82 S.Ct. 468, 7 L.Ed.2d 417 (1962) ; Hayman , 342 U.S. at 219, 72 S.Ct. 263, but permitting federal prisoners to file habeas petitions based on an intervening change in statutory interpretation provides those prisoners with a superior remedy. Allowing a prisoner to use the saving clause to bring a statutory claim in a habeas petition circumvents the bar on successive petitions. 28 U.S.C. § 2255(h). It does away with the one-year statute of limitations. Id. § 2255(f). It renders the process for obtaining permission to file a second or successive motion, id. § 2253(b), and that for obtaining a certificate of appealability, id. § 2253(c)(1), a nullity. A prisoner who brings a constitutional claim under section 2255(h), in contrast, must overcome these procedural hurdles. The Wofford test unravels this carefully tailored scheme. It makes no sense to allow a federal prisoner to evade the statutory framework by filing a petition for a writ of habeas corpus.

Several of the separate opinions raise a version of the argument that a previously adequate remedy may later become inadequate, but these temporal arguments fail in the light of the whole text. Judge Rosenbaum's dissent states that "as a practical matter" a "right ... cannot be vindicated until after the Supreme Court announces the new rule." Rosenbaum Dissent at 1154. But this argument ignores that litigants often make novel arguments in the hope that a court will adopt them as a matter of first impression or in a rejection of past precedent. Judge Martin's dissent and Judge Jordan's concurring opinion argue that the "present tense" of the saving clause requires that we ask whether section 2255"is adequate or ineffective to test" "at the time the petition is filed in federal court." Martin Dissent at 1115; Jordan Concurring at 1104 (We "assess inadequacy and ineffectiveness as of the time [a petitioner] files his § 2241 habeas corpus petition, and not as of the time when he submitted his initial § 2255 motion."). But whether the remedy "is" inadequate or ineffective must refer to the nature of the remedy, not to one specific motion, or else the motion becomes inadequate every time a procedural rule like the statute of limitations or procedural default prevents success. The procedural bars mean nothing if they can be avoided through the saving clause. The saving clause does not allow access to section 2241 whenever a claim is untimely or procedurally defaulted otherwise the statute would render itself inadequate or ineffective. The same must be true for the bar on second or successive motions. Contrary to Judge Martin's dissent, Martin Dissent at 1117–19, the means also do not somehow become inadequate or ineffective when circuit precedent is abrogated after a prisoner has filed his first motion to vacate. When Congress limits a prisoner to a single motion to vacate, it does not render the "remedy by motion inadequate or ineffective to test the legality of his detention," 28 U.S.C. § 2255(e) ; it instead limits each prisoner to one test.

Allowing a federal prisoner to bring a successive claim in a petition for a writ of habeas corpus also defies the logic of the venue provisions. A federal prisoner must file a motion to vacate in the court that tried and sentenced him, where he can challenge issues about his trial and sentencing. See id. § 2255(a). In contrast, he must bring a petition for a writ of habeas corpus in the district in which he is imprisoned, where he can challenge his detention. See id. § 2241(d). The United States Attorney who participated in sentencing defends challenges to the prisoner's trial and sentencing. Id. § 2255(a). But the warden of the prison defends challenges to the prisoner's detention. Id. § 2241(d).

Allowing a prisoner to bring an ordinary attack on his sentence in the district where he is detained eviscerates this structure. It resurrects the problems that section 2255 was enacted to solve, such as heavy burdens on courts located in districts with federal prisons, inconvenience for witnesses who must travel far from where the prisoner was tried to the place where he is detained, and the requirement that wardens defend resentencing. See Hayman , 342 U.S. at 219, 213, 72 S.Ct. 263. It also creates new procedural and jurisdictional wrinkles for district courts tasked with implementing relief that the statute does not contemplate. See Hill v. Sepanek , Civil No. 14-85-ART, 2017 WL 73338, at *5–9 (E.D. Ky. Jan. 6, 2017) (Thapar, J.) ("[P]ractical problems ... arise under any construction of the saving[ ] clause that does not comport with its plain meaning."); Love v. Hogsten , Civil Action No. 1:09-cv-2134-JEC, 2012 WL 3822194, at *4 (N.D. Ga. Sept. 4, 2012) (J. Carnes, J.) ("Insisting that what is essentially a § 2255 claim ... be instead deemed a § 2241 claim [shifts] the venue ... from the district of sentencing to the district in which the petitioner is confined[,] ... meaning that there is the potential for multiple § 2241 saving[ ] clause claims in multiple districts, creating confusion, duplicative effort, and potentially inconsistent results."). Allowing access to the saving clause to bring ordinary sentencing challenges disregards Congress's decision to bifurcate the system of collateral review between challenges to a prisoner's sentence and challenges to the execution of a prisoner's sentence. Limiting the saving clause to claims that are not cognizable or that cannot be remedied under section 2255 respects the entire system of federal collateral review.

The government and some of the separate opinions argue that our interpretation renders the saving clause meaningless, see Jordan Concurring at 1111; Martin Dissent at 1119, but we disagree. The saving clause has meaning because not all claims can be remedied by section 2255. A prisoner sentenced by a federal court, for example, may file a petition for a writ of habeas corpus to challenge the execution of his sentence, such as the deprivation of good-time credits or parole determinations. See, e.g., Hajduk , 764 F.2d at 796. The saving clause also allows a prisoner to bring a petition for a writ of habeas corpus when the sentencing court is unavailable. Other circuits have held that a prisoner may file a petition for a writ of habeas corpus if his sentencing court has been dissolved. See Prost , 636 F.3d at 588 (explaining that for military prisoners "the resort to § 2241 is the norm rather than the exception ... due to the evanescent nature of court martial proceedings: the sentencing court literally dissolves after sentencing and is no longer available to test a prisoner's collateral attack"). Or, as our sister circuit has held, perhaps practical considerations (such as multiple sentencing courts) might prevent a petitioner from filing a motion to vacate. See Cohen v. United States , 593 F.2d 766, 771 & n.12 (6th Cir. 1979). "But only in those kinds of limited circumstances is [the remedy by motion] 'inadequate or ineffective to test the legality of his detention.' " Samak , 766 F.3d at 1278 (W. Pryor, J., concurring) (quoting 28 U.S.C. § 2255(e) ).

Judge Rosenbaum's dissent and Judge Jordan's concurring opinion argue that our interpretation conflicts with the text of the statute because, in their view, a prisoner can petition for a writ of habeas corpus to challenge the execution of his sentence without accessing the saving clause. Rosenbaum Dissent at 1124–29; Jordan Concurring at 1109. But they misinterpret "a prisoner who is authorized to apply for relief by motion pursuant to this section," id. § 2255(e), to mean only a prisoner bringing a claim under section 2255(a). Rosenbaum Dissent at 1124–29; Jordan Concurring at 1108. The use of the broader word "detention" suggests that the saving clause applies to claims about the execution of a sentence because we would expect the clause to say "sentence" if it only applied to sentencing claims. And the phrase "a prisoner who" highlights that the prohibition on petitioning for a writ of habeas corpus applies to a kind of person, namely "a prisoner in custody under sentence of a court established by Act of Congress," id. § 2255(a), not a kind of claim. The better interpretation of "pursuant to this section" is in opposition to prisoners authorized pursuant to a different section, such as "a person in custody pursuant to the judgment of a State court" in the neighboring section of the code, id. § 2254(a).

Despite the dissent's attempt to limit the meaning of "authorized" to portions of section 2255 that contain affirmative grants to the prisoner, as opposed to processing instructions to the court, Rosenbaum Dissent at 1124–27, the text will not bear this interpretation. The phrase "pursuant to this section" refers to all of section 2255, not the first subsection alone. An ordinary speaker of English would not understand a prisoner who lacks permission to file a second or successive motion to be any more "authorized to apply for relief" than a prisoner with a claim outside of section 2255(a). But if that prisoner can instead petition for a writ of habeas corpus, then section 2255(h) becomes a nullity. Our interpretation that a prisoner is "authorized to apply for relief by motion pursuant to this section," id. § 2255(e), if he is "in custody under sentence of a court established by Act of Congress," id. § 2255(a), avoids this nullity. But most importantly, even if a prisoner with a claim based on the execution of his sentence could petition for a writ of habeas corpus without the saving clause, the saving clause would still apply to situations in which a federal sentencing court dissolves or access to the remedy by motion is impractical. The interpretation presented by Judge Rosenbaum's dissent and Judge Jordan's concurring opinion proves nothing about whether a prisoner with a claim based on a change in caselaw or a prisoner with a claim based on actual innocence satisfies the saving clause.

Judge Rosenbaum's dissent contends that our interpretation of the saving clause violates the Suspension Clause, U.S. Const. Art. 1, § 9, cl. 2, Rosenbaum Dissent at 1132–34, but we disagree. We have no need to use the saving clause as a fount of constitutional avoidance in this appeal, see Rosenbaum Dissent at 1123, 1135–36, because there is no constitutional violation to avoid. The Suspension Clause provides that "[t]he Privilege of the Writ of Habeas Corpus shall not be suspended, unless when in Cases of Rebellion or Invasion the public Safety may require it." U.S. Const. Art. 1, § 9, cl. 2. Our interpretation of the saving clause cannot suspend the writ because the Original Writ in the Supreme Court remains available, habeas corpus at common law did not apply to prisoners sentenced by a court of competent jurisdiction, and the decision of the Supreme Court in Felker v. Turpin , 518 U.S. 651, 116 S.Ct. 2333, 135 L.Ed.2d 827 (1996), upheld a bar on successive motions against constitutional challenge.

Nothing in the Antiterrorism and Effective Death Penalty Act "strangle [s] the power of the Supreme Court to grant an Original Writ." Samak , 766 F.3d at 1291 (W. Pryor, J., concurring) (citing Felker , 518 U.S. at 658, 116 S.Ct. 2333 ). "The Act cannot transgress the constitutional rights of prisoners who allege that they have been erroneously sentenced or unfairly tried when the Supreme Court retains its power to grant an Original Writ. The Supreme Court affirmed this proposition as early as 1868 ... and as recently as 1996." Id. at 1292 (citing Ex Parte Yerger , 75 U.S. (8 Wall.) 85, 105, 19 L.Ed. 332 (1868) ; Felker , 518 U.S. at 651, 116 S.Ct. 2333 ). Judge Rosenbaum's dissent argues that the Original Writ is not an adequate substitute for the common law writ because it would be impractical and the Supreme Court rarely grants the writ. Rosenbaum Dissent at 1146–47. But "judgments about the proper scope of the writ are 'normally for Congress to make.' " Felker , 518 U.S. at 664, 116 S.Ct. 2333 (citing Lonchar v. Thomas , 517 U.S. 314, 323, 116 S.Ct. 1293, 134 L.Ed.2d 440 (1996) ). And because the Constitution does not even require Congress to create inferior courts, U.S. Const. Art. III, § 1, it makes no sense to assert that a remedy within the original jurisdiction of the Supreme Court is insufficient to satisfy the Suspension Clause.

To argue that section 2255 suspends the writ ignores that at common law, the writ of habeas corpus would not have been available at all to prisoners like McCarthan. And the Supreme Court has never held that the constitutional requirements of the Suspension Clause increase over time. See I.N.S. v. St. Cyr , 533 U.S. 289, 300–301, 121 S.Ct. 2271, 150 L.Ed.2d 347 (2001) ; Felker , 518 U.S. at 663–64, 116 S.Ct. 2333. Traditional habeas corpus dealt with only "serious abuses of power by a government, say a king's imprisonment of an individual without referring the matter to a court." Lonchar , 517 U.S. at 322, 116 S.Ct. 1293. "As limited by the act of 1789, [the writ] did not extend to cases of imprisonment after conviction, under sentences of competent tribunals...." Ex Parte Yerger , 75 U.S. (8 Wall.) at 101. Judge Rosenbaum's dissent relies heavily on Boumediene v. Bush , 553 U.S. 723, 128 S.Ct. 2229, 171 L.Ed.2d 41 (2008), but this reliance is misplaced. In Boumediene , the Supreme Court addressed the scope of habeas corpus for executive detainees "where no trial has been held" and distinguished decisions like Felker in which a prisoner sought relief from a judgment imposed in a "fair, adversary proceeding." Id. at 732, 774, 782, 128 S.Ct. 2229. Because McCarthan would not have had the right to habeas corpus under the common law, his inability to file a second collateral attack after a change in caselaw cannot possibly constitute a suspension of the writ.

Judge Rosenbaum's dissent argues that the Suspension Clause requires the availability of successive petitions for new rules of statutory interpretation, Rosenbaum Dissent at 1122–23, but as the decision of the Supreme Court in Felker reminds us, the writ has not been suspended whenever a prisoner cannot file a successive collateral attack. The Antiterrorism and Effective Death Penalty Act creates parallel procedures for federal and state prisoners: federal prisoners bring collateral attacks under section 2255, and state prisoners bring collateral attacks under section 2254. Both remedies include a nearly identical bar on successive attacks, 28 U.S.C. §§ 2244(b), 2255(h), but only the federal remedy includes a saving clause. In Felker , the Supreme Court held that the bar on second or successive collateral attacks by state prisoners did not violate the Suspension Clause. 518 U.S. at 664, 116 S.Ct. 2333. Because the Supreme Court has approved limitations on successive petitions without a saving clause, those same limitations with a saving clause must be constitutional. Citing the separation of powers to limit the application of Felker to state prisoners, Rosenbaum Dissent at 1143–45, is "interpretive jiggery-pokery," King v. Burwell, ––– U.S. ––––, 135 S.Ct. 2480, 2500, 192 L.Ed.2d 483 (2015) (Scalia, J., dissenting), that conveniently ignores the dissent's own insistence that "limited government powers" also animate habeas corpus, Rosenbaum Dissent at 1122.

Contrary to Judge Rosenbaum's dissent, retroactivity doctrine does not undermine this conclusion; indeed, it is unclear why retroactivity is even relevant. When the Supreme Court makes a right retroactively available on collateral review, it does not mean that a prisoner is constitutionally entitled to have a court review a violation of that right on the merits. See, e.g. , Bousley v. United States , 523 U.S. 614, 621, 118 S.Ct. 1604, 140 L.Ed.2d 828 (1998) ("Though petitioner's claim is [retroactive], there are nonetheless significant procedural hurdles to its consideration on the merits."). Retroactivity means that a court is no longer barred from applying a new rule on collateral review, not that a court must create a vehicle for collateral review because there is a new rule. Procedural barriers like procedural default, id. , the statute of limitations, or the bar on successive motions may prevent litigation about a violation of that new rule. That a procedural rule prevents litigating an error does not create a constitutional crisis, let alone a suspension of the writ. See, e.g. , Felker , 518 U.S. at 664, 116 S.Ct. 2333 ("The added restrictions which the Act places on second habeas petitions ... do not amount to a 'suspension' of the writ."). Our current, erroneous precedent is not dictated by constitutional avoidance concerns. If anything, we conform our precedent to the Constitution by rejecting an atextual judicial invention and faithfully interpreting the text of the statute.

C. Precedent and Stare Decisis

We recognize that overturning precedent is and should be a rare occurrence. Our Court follows the principles of stare decisis as described by the Supreme Court. See Glazner v. Glazner , 347 F.3d 1212, 1216 (11th Cir. 2003) (en banc). Courts "should not lightly overrule past decisions," Moragne v. States Marine Lines, Inc. , 398 U.S. 375, 403, 90 S.Ct. 1772, 26 L.Ed.2d 339 (1970), because "[s]tability and predictability are essential factors in the proper operation of the rule of law," Bonner v. City of Prichard , 661 F.2d 1206, 1209 (11th Cir. 1981). Stare decisis is especially important when we construe statutes because "Congress remains free to alter what we have done." John R. Sand & Gravel Co. v. United States , 552 U.S. 130, 139, 128 S.Ct. 750, 169 L.Ed.2d 591 (2008) (quoting Patterson v. McLean Credit Union , 491 U.S. 164, 172–73, 109 S.Ct. 2363, 105 L.Ed.2d 132 (1989) ).

But stare decisis "is not an inexorable command." Payne v. Tennessee , 501 U.S. 808, 828, 111 S.Ct. 2597, 115 L.Ed.2d 720 (1991) ; accord Bryan A. Garner, et al. , The Law of Judicial Precedent 388 (2016) ("[S]tare decisis isn't an ineluctable doctrine to be applied with procrustean rigor."). We may overrule precedent that is "plainly and palpably wrong" if overruling the precedent would not "result in more harm than continuing to follow the erroneous decision." Garner, et al. , supra , at 388. Our Court has held that "we must follow the Supreme Court's instruction that stare decisis should be abandoned where, as here, 'a prior judicial ruling should come to be seen so clearly as error that its enforcement was for that very reason doomed.' " Glazner , 347 F.3d at 1216 (quoting Planned Parenthood of Se. Pa. v. Casey , 505 U.S. 833, 854, 112 S.Ct. 2791, 120 L.Ed.2d 674 (1992) ). If the statutory and doctrinal underpinnings have eroded and there has not been significant reliance on the precedent, it may be abandoned. Kimble v. Marvel Entm't, LLC , ––– U.S. ––––, 135 S.Ct. 2401, 2410–11, 192 L.Ed.2d 463 (2015). The same is true if a decision has "proved unworkable." Id.

In this instance, we take the rare step of overruling our precedents for three reasons. First, they are wholly divorced from the text. Second, reliance interests are minimal. And third, our precedents have proved unworkable. Continuing to follow these erroneous precedents would do more harm than good.

First, our precedents are not faithful to the text of the statute. As discussed above, the Wofford test is "plainly and palpably wrong." See Garner, et al. , supra , at 388. Even our colleagues writing separately agree that our precedents are wrong. Rosenbaum Dissent at 1121 ("I agree with the Majority that we incorrectly interpreted 28 U.S.C. § 2255(e) on at least five occasions."); Martin Dissent at 1114 ("I have always believed that Wofford was wrong and that this Court's rulings on saving[ ] clause cases that have since followed Wofford are wrong as well."); Jordan Concurring at 1101 ("I agree with the majority's ultimate conclusion that the 'saving clause' of § 2255(e) does not permit sentencing claims ... [b]ut my reading of the 'saving clause' is broader."); Wilson Dissent at 1111 (agreeing with Judge Jordan's textual analysis but extending it to a prisoner whose sentence exceeds the statutory maximum). We have previously decided to overturn our precedent when the statute is "clear and unambiguous," Glazner , 347 F.3d at 1215, or the precedent is "inconsistent" with the text of the statute, United States v. Svete , 556 F.3d 1157, 1166 (11th Cir. 2009) (en banc). Our decisions in Wofford , Gilbert , Williams , and Bryant ignored the text. When we first addressed the saving clause, "we went straight to the legislative history of the clause to divine its meaning, but unsurprisingly could find no clues." Samak , 766 F.3d at 1276 (W. Pryor, J., concurring). And as we applied this rule, we never returned to a careful consideration of the text. See Williams , 713 F.3d at 1341 ("[T]he statute says precious little about what it means for the original motion to have been 'inadequate' or 'ineffective.' "); Gilbert , 640 F.3d at 1307 ("This is one of those times when it is easier to determine something that a provision does not mean...."). Because our precedents are so far removed from the text of the statute, there is less reason to defer to them.

Nor is there a settled consensus about the meaning of the saving clause. The Tenth Circuit adheres to the text, Prost , 636 F.3d at 584–87, but most of our sister circuits have focused on legislative purpose and avoided rigorous textual analysis. As a result, a wide variety of interpretations of the saving clause exists amongst the circuits. For example, the Fifth Circuit has refused to apply the saving clause to sentencing errors, In re Bradford , 660 F.3d 226, 230 (5th Cir. 2011), our Circuit has extended the saving clause only to sentencing errors that exceed the statutory maximum, Bryant , 738 F.3d at 1274, and the Seventh Circuit has extended the saving clause to all sentencing errors, including those under the then-mandatory sentencing guidelines, Brown , 719 F.3d at 587. Although several circuits have adopted some version of a "circuit busting" test, they do not agree on its contours. Some require actual innocence, see, e.g. , Wooten , 677 F.3d at 307–08, others require "complete miscarriage of justice," In re Dorsainvil , 119 F.3d at 251, and others focus on constitutional avoidance, Triestman , 124 F.3d at 376. And our dissenting colleagues propose new tests based on constitutional avoidance, Rosenbaum Dissent at 1121–22, or fundamental defect, Martin Dissent at 1116. This is not a situation where our precedents align with a uniform interpretation.

Second, reliance interests for our precedents are minimal. As a fundamental matter, rules about collateral review do not create significant reliance interests. In areas of law like property and contracts, reliance interests are particularly strong. Kimble , 135 S.Ct. at 2410. And of course, individuals rely on criminal law because it regulates primary conduct. But unlike rules of property, where court decisions are "retrospective and may affect titles purchased on the faith of their stability," Garner, et al. , supra , at 422 (quoting Minn. Mining Co. v. Nat'l Mining Co. , 70 U.S. (3 Wall.) 332, 334, 18 L.Ed. 42 (1865) ), the availability of collateral review does not prompt reliance. Whether a federal prisoner receives one round of collateral review or two does not impact his decision-making. He cannot rely on an unanticipated change in law. And if he intentionally withheld an argument for a second bite at the apple, he would not have preserved it.

Our current test is relatively recent and has rarely led to a grant of relief. Although we have been applying some version of the test for intervening changes in law since Wofford , we did not grant relief until Bryant . When we considered the matter en banc in Gilbert , we still referred to the Wofford test as "dicta." 640 F.3d at 1319. And it has been used to grant relief only one other time, in Mackey . "[G]overning decisions" are more easily overturned "if the precedent is particularly recent and has not generated any serious reliance interests" or if it has "sustained serious erosion from our recent decisions." Al-Sharif v. U.S. Citizenship and Immigr. Servs. , 734 F.3d 207, 212 (3d Cir. 2013) (quotation marks and citations omitted).

There is not even a clear consensus about what the Wofford test entails, even among the parties to this appeal who favor its retention. McCarthan argues in favor of the Wofford test as applied in Bryant but quibbles over the meaning of step four. The warden agrees that the Wofford test is correct but describes its analysis as a three-part test: "(1) a Supreme Court decision of statutory construction has changed controlling circuit law retroactively; (2) in a way that establishes a fundamental defect in the prisoner's conviction or sentence and renders his continued detention illegal; and (3) the prisoner had no reasonable opportunity for a judicial remedy of that fundamental defect in another proceeding." And the National Association of Criminal Defense Lawyers as amicus curiae first describes Wofford as a four-factor test, and then proposes a new test: "whether the prisoner had a genuine opportunity to raise his claim in an adequate and effective fashion." This cacophony highlights that we do not face a problem of overturning a long-established, settled test. In fact, it is difficult to imagine how someone could rely on a test so inscrutable. And when reliance interests are minimal, a court may overrule its own precedent. Garner, et al. , supra , at 401, 408–09.

Third, our precedents have proved unworkable. Wofford has placed a heavy burden on courts in this Circuit. Hundreds, perhaps thousands, of prisoners have filed petitions citing the Wofford test in the various districts where federal prisons are located in this Circuit. These doomed collateral attacks have required wardens to defend decades-old sentencing determinations and resurrected the exact problems that Congress attempted to solve when it created the remedy by motion. Hayman , 342 U.S. at 219, 72 S.Ct. 263.

And the Wofford test is burdensome to apply. For example, deciding whether circuit precedent foreclosed an argument, whether the Supreme Court abrogated that precedent, and whether that decision applies retroactively on collateral review can be a difficult and controversial task. See, e.g. , McCarthan , 811 F.3d at 1247–50 (discussing the argument that circuit precedent about a Georgia statute did not foreclose McCarthan from challenging a Florida statute); In re Watkins , 810 F.3d 375, 380 (6th Cir. 2015) (describing the circuit split regarding whether Johnson v. United States , ––– U.S. ––––, 135 S.Ct. 2551, 192 L.Ed.2d 569 (2015), applies retroactively on collateral review). In this appeal, the panel disagreed about how to apply the Wofford test to McCarthan. The separate concurring opinion stated that "the cumbersome nature of that test leads to just the type of confusion we have here," namely that the steps of the analysis overlap. McCarthan , 811 F.3d at 1257 (Proctor, J., concurring). In Samak , we had to apply the law of another circuit to answer these questions because a federal inmate imprisoned in our circuit was sentenced in another. 766 F.3d at 1275 n.3. In Cortes-Morales , we were asked to decide whether the logic of the Wofford test should extend to retroactive amendments to state legislation. 827 F.3d at 1016. Hard questions with no predictable answers will continue to arise if we insist on applying a test unbound by the text.

Even the questions of the Wofford test that seem straightforward prove difficult. The Wofford test requires that we determine whether the current sentence exceeds the statutory maximum. The panel opinion in this appeal went to great lengths to distinguish this inquiry from the merits of the motion to vacate. The difference, according to the panel, is that the Wofford test looks at both "invalid predicate convictions that a federal prisoner could not have challenged in his initial § 2255 petition because any challenge was squarely foreclosed by binding Circuit precedent that the Supreme Court only subsequently overturned ('squarely foreclosed convictions')" and "invalid predicate convictions that a defendant could have, but failed to, challenge earlier ('erroneously counted convictions')," but the merits inquiry looks only at the former. McCarthan , 811 F.3d at 1251. In the final part of the Wofford test, we are expected to resolve "whether the saving[ ] clause in § 2255(e) reaches" the kind of claim presented, Bryant , 738 F.3d at 1274, but this circular standard is meaningless when the issue is one of first impression.

All of these difficult and convoluted determinations are made in a threshold jurisdictional analysis. The labyrinthian analysis required by our precedents is not a prudent use of judicial resources. See Duckworth , 454 U.S. at 4, 102 S.Ct. 18 ("Creating a new exception" means that "[s]ignificantly more time and resources would be consumed as district and appellate courts examined the merits to determine whether a claim met the requisite level of validity."). True, it is often "more important that the applicable rule of law be settled than that it be settled right." Agostini v. Felton , 521 U.S. 203, 235, 117 S.Ct. 1997, 138 L.Ed.2d 391 1997 (quoting Burnet v. Coronado Oil & Gas Co. , 285 U.S. 393, 406, 52 S.Ct. 443, 76 L.Ed. 815 (1932) (Brandeis, J., dissenting)). But the Wofford test repeatedly unsettles the law in this Circuit.

Because the Wofford test ignores the text of the saving clause, induces no reliance, and burdens our courts, it does more harm than good. "[T]his weighing of alternative harms is the normal assessment in deciding whether to overrule precedent." Garner, et al. , supra , at 388. In contrast, being faithful to the text of the saving clause makes our task simple, predictable, and sensible. This appeal presents the rare circumstance where we should overturn our precedents.

Contrary to McCarthan's argument, this appeal presents no problems of justiciability. McCarthan argues that because both parties accept the Wofford test and "[n]either party stands to obtain meaningful relief from a re-consideration of [it]," this Court lacks jurisdiction to reconsider its precedents. See U.S. Const. Art. III, § 2, cl. 1. Nonsense. As the Warden correctly responds, there is a live case or controversy about whether McCarthan is entitled to relief, "notwithstanding any agreement on the rules to be applied." We have a responsibility to interpret the law correctly. And the Supreme Court has modeled the use of an amicus curiae to aid in this endeavor. SeeMiscellaneous Order , Irizarry v. United States , 552 U.S. 1135, 128 S.Ct. 977, 169 L.Ed.2d 799 (2008) (inviting Peter B. Rutledge to brief and argue the case, as amicus curiae , in support of the judgment below); Miscellaneous Order , Beckles v. United States , ––– U.S. ––––, 137 S.Ct. 23, 195 L.Ed.2d 895 (2016) (inviting Adam K. Mortara to brief and argue the case, as amicus curiae , in support of the judgment below). We must interpret the statute that governs this appeal and apply it to the parties before us. We have done so.

A motion to vacate is inadequate or ineffective to test the legality of a prisoner's detention only when it cannot remedy a particular kind of claim. Even if a prisoner's claim fails under circuit precedent, a motion to vacate remains an adequate and effective remedy for a prisoner to raise the claim and attempt to persuade the court to change its precedent, and failing that, to seek certiorari in the Supreme Court. McCarthan does not qualify for the saving clause because his claim that escape is not a violent felony is cognizable under section 2255. Because he "was free to bring" this claim about the interpretation of his sentencing law in his initial motion to vacate, the remedy by motion was an "adequate and effective means for testing such an argument." Prost , 636 F.3d at 580. He cannot now use the saving clause to make that claim in a petition for a writ of habeas corpus.

IV. CONCLUSION

For all the above reasons, we overrule the Wofford test as applied in Bryant and Mackey and AFFIRM the order denying McCarthan's petition for a writ of habeas corpus.

ED CARNES, Chief Judge, concurring:

I join in full the opinion of the Court and write separately only to emphasize a point that it makes in passing.

Judge Rosenbaum's dissenting opinion says, in effect, that there should be another exception to the bar on second and successive § 2255 motions to permit claims based on a new decision about the scope of a criminal statute that the Supreme Court has made retroactively applicable to cases on first collateral review. But Congress has not said that. The place to have said it, of course, would have been in § 2255(h), which contains the two exceptions to the bar on second and successive motions. Congress could have simply added a third exception to the list so that subsection (h) would have read:

A second or successive motion must be certified as provided in section 2244 by a panel of the appropriate court of appeals to contain—

(1) newly discovered evidence that, if proven and viewed in light of the evidence as a whole, would be sufficient to establish by clear and convincing evidence that no reasonable factfinder would have found the movant guilty of the offense; or

(2) a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable, or

(3) a new rule of statutory law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable.

Simple as that would have been, Congress did not do it. Instead, it limited the exceptions to two.

The dissenting opinion would have us "improve" the statute by writing in the exception that it favors, but we cannot do that. As the Supreme Court has instructed us: "It is for Congress, not this Court, to amend the statute if it believes that the interplay of [some provisions] of § 2255 unduly restricts federal prisoners' ability to file second or successive motions." Dodd v. United States , 545 U.S. 353, 359–60, 125 S.Ct. 2478, 2483, 162 L.Ed.2d 343 (2005).

There is more at stake here than an issue of statutory interpretation. The question is one of the proper role of the judiciary. As we have explained, "Our oft-stated rule against judicial revision of statutes finds plenty of anchor weight in the bedrock principle that we are a country of laws, not one ruled by the musings, whether pragmatic or otherwise, of the black-robed class." T-Mobile S., LLC v. City of Milton, Ga. , 728 F.3d 1274, 1285 (11th Cir. 2013) ; see also id. at 1284 ("We are interpreting a statute, not designing one. Although we, like most judges, have enough ego to believe that we could improve a good many statutes if given the chance, statutory construction does not give us that chance if we are true to the judicial function. Our duty is to say what statutory language means, not what it should mean, and not what it would mean if we had drafted it."); Friends of the Everglades v. S. Fla. Water Mgmt. Dist. , 570 F.3d 1210, 1224 (11th Cir. 2009) ("[W]e are not allowed to add or subtract words from a statute; we cannot rewrite it."); Wright v. Sec'y for Dep't of Corr. , 278 F.3d 1245, 1255 (11th Cir. 2002) ("Our function is to apply statutes, to carry out the expression of the legislative will that is embodied in them, not to 'improve' statutes by altering them."); Harris v. Garner , 216 F.3d 970, 976 (11th Cir. 2000) (en banc) ("We will not do to the statutory language what Congress did not do with it, because the role of the judicial branch is to apply statutory language, not to rewrite it.").

The dissent invokes the separation of powers, but that constitutional doctrine is best served by respecting the fundamental principle that it is the role of Congress, not the Courts, to decide what the statutory law is to be, and Congress has done that in § 2255(h). We honor the separation of powers doctrine when we resist the temptation, irresistible as it may seem, to judicially revise statutes to suit our sense of sound policy.

The dissenting opinion repeatedly protests that it is not an example of "judicial activism," using that phrase more than a dozen times to answer an unstated charge. And with painful accuracy, that opinion charges me with being among the judges who have attempted to improve the saving clause of § 2255(e) by interpretation. See Rosenbaum Dissent at 1157. Mea culpa. As the author of the Wofford opinion and its dicta, I am the one who laid out the seedbed from which the weeds have grown around this issue in our circuit. Having to watch for the past 17 years as my woefully wrong Wofford opinion worked its mischief is the price I have paid for my sin, or at least for that particular one. And no one knows sin like an old sinner.

Which brings to mind the various formulations that other judges have used to admit their mistakes. Nearly everyone's favorite is Justice Frankfurter's: "Wisdom too often never comes, and so one ought not to reject it merely because it comes late." Henslee v. Union Planters Nat. Bank & Trust Co. , 335 U.S. 595, 600, 69 S.Ct. 290, 293, 93 L.Ed. 259 (1949) (Frankfurter, J., dissenting). But there are other ways of phrasing judicial repentance. See, e.g. , Dart Cherokee Basin Operating Co. v. Owens , 574 U.S. ––––, 135 S.Ct. 547, 561, 190 L.Ed.2d 495 (2014) (Scalia, J., dissenting) ("As for my own culpability in overlooking the issue, I must accept that and will take it with me to the grave."); Massachusetts v. United States , 333 U.S. 611, 639–40, 68 S.Ct. 747, 763, 92 L.Ed. 968 (1948) (Jackson, J., dissenting) ("I see no reason why I should be consciously wrong today because I was unconsciously wrong yesterday."). I prefer to put it more colloquially: Wofford was a screw up. To repeat the error by revising that opinion's revision of the saving clause in another attempt to improve the text of the statute would be another screw up. Once is enough for me.

JORDAN, Circuit Judge, concurring in part and dissenting in part.

The text of 28 U.S.C. § 2255(e) has remained unchanged since 1948, despite Congress' significant overhaul of federal collateral review in 1996. Given the difficult task of deciphering language designed for a bygone era in a post-AEDPA world, it is no wonder that federal courts have struggled to reach a uniform understanding. Recognizing that the meaning of § 2255(e)"is not easy of solution," United States v. Universal C.I.T. Credit Corp. , 344 U.S. 218, 221, 73 S.Ct. 227, 97 L.Ed. 260 (1952), I offer my own perspective.

I agree with the majority's ultimate conclusion that the "saving clause" of § 2255(e) does not permit sentencing claims like the one asserted by Mr. McCarthan. But my reading of the "saving clause" is broader than the one articulated by the majority. In my view, the "saving clause" allows a federal prisoner to seek a writ of habeas corpus pursuant to 28 U.S.C. § 2241 if § 2255 relief is unavailable to him and a new (and governing) interpretation of the statute of conviction demonstrates that he never committed a crime. Because the majority's reading of § 2255(e) apparently forecloses a habeas remedy in such circumstances, I concur only in the judgment.

I

Suppose that Congress enacts a new criminal statute, 18 U.S.C. § 999.99, and makes a violation of that statute punishable by up to ten years in prison. The Department of Justice believes that persons who commit acts A and B violate § 999.99, and sets out to prosecute persons who have committed those acts.

A federal grand jury in the Southern District of Florida indicts Joe Unlucky for violating § 999.99 by committing acts A and B. At trial, Mr. Unlucky argues that, when read properly, the statute does not criminalize acts A and B. The district court rejects the argument and instructs the jury that it may return a verdict of guilty if it finds that Mr. Unlucky committed acts A and B. Because the government puts on undisputed evidence that Mr. Unlucky did in fact commit acts A and B, the jury finds him guilty, and the district court sentences him to eight years in prison. On appeal, the Eleventh Circuit affirms, and rejects Mr. Unlucky's reading of the statute. The Supreme Court denies certiorari.

Mr. Unlucky then files a timely motion to vacate pursuant to 28 U.S.C. § 2255. Because his contention about the scope of § 999.99 has already been rejected on direct appeal, and there have been no intervening changes in governing law, he is not able to reassert the same claim again. See generally Rozier v. United States , 701 F.3d 681, 684 (11th Cir. 2012) ("At least where there has been no intervening change in controlling law, a claim or issue that was decided against a defendant on direct appeal may not be the basis for relief in a § 2255 proceeding."). So Mr. Unlucky alleges in his motion that his trial counsel rendered ineffective assistance at sentencing. The district court holds an evidentiary hearing, rejects the ineffectiveness claim on the merits, and grants a certificate of appealability. The Eleventh Circuit subsequently affirms the district court's denial of Mr. Unlucky's motion to vacate, and the Supreme Court again denies certiorari.

Two years later, the Supreme Court decides a case just like Mr. Unlucky's, and holds that a person who commits acts A and B does not violate § 999.99. This decision by the Supreme Court, of course, means that § 999.99 never criminalized acts A and B, for "[a] judicial construction of a statute is an authoritative statement of what the statute meant before as well as after the decision of the case giving rise to that construction." Rivers v. Roadway Express, Inc. , 511 U.S. 298, 312–13, 114 S.Ct. 1510, 128 L.Ed.2d 274 (1994). The decision, moreover, is fully retroactive to cases on collateral review under Bousley v. United States , 523 U.S. 614, 620, 118 S.Ct. 1604, 140 L.Ed.2d 828 (1998).

So it turns out that Mr. Unlucky never committed the federal offense with which he was charged and for which he was convicted. Yet he sits in a federal prison with about four years left to serve on his sentence for a non-existent crime.

Mr. Unlucky thinks about filing another § 2255 motion based on the new Supreme Court decision, but quickly finds out that he cannot do so. The problem is that AEDPA permits a second or successive motion to vacate only where newly discovered evidence establishes "by clear and convincing evidence that no reasonable factfinder would have found the [person] guilty of the offense," or where there is a "new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable." See 28 U.S.C. § 2255(h)(1)–(2). Mr. Unlucky does not have any newly discovered evidence, and the new Supreme Court decision interpreting § 999.99 is statutory, not constitutional. See In re Thomas , 823 F.3d 1345, 1349 (11th Cir. 2016) ("We have held that [the] Supreme Court has not announced a new rule of constitutional law when it has merely interpreted an existing statute."). He therefore cannot satisfy the requirements of § 2255(h) for filing a second or successive motion to vacate.

Not wanting to waste more years of his life in prison for a non-existent crime, Mr. Unlucky, relying on the "saving clause" of § 2255(e), files a habeas corpus petition pursuant to § 2241. He requests that the district court vacate his conviction based on the new Supreme Court decision and order his release from custody.

In this setting, can Mr. Unlucky use the "saving clause" to seek habeas corpus relief? I think so.

II

Comprised of a single, bedeviling sentence with various clauses, § 2255(e) states as follows:

[1] An application for a writ of habeas corpus in behalf of a prisoner who is authorized to apply for relief by motion pursuant to this section, shall not be entertained if it appears that the applicant has failed to apply for relief, by motion, to the court which sentenced him, or that such court has denied him relief, [2] unless it also appears that the remedy by motion is inadequate or ineffective to test the legality of his detention.

(brackets added).

The first clause of § 2255(e) is the "authorization clause," and the second clause—the one that has proven most difficult to figure out—is the "saving clause." As I hope to explain, Mr. Unlucky can use the "saving clause" to file a habeas corpus petition because a "remedy by [§ 2255 ] motion is inadequate or ineffective to test the legality of his detention."

A

"Congress' use of a verb tense is significant in construing statutes," United States v. Wilson , 503 U.S. 329, 333, 112 S.Ct. 1351, 117 L.Ed.2d 593 (1992), so I begin with the verbs that Congress chose to use in § 2255(e). As Judge Martin correctly points out in her dissenting opinion, see Martin Dissent at 1114–15, Congress' choice of the word "is" in the "saving clause" is significant.

The "authorization clause" twice employs the past tense ("has failed" and "has denied"), while the "saving clause" uses a single present-tense verb ("is") right before the words "inadequate or ineffective." Because Congress used a present-tense verb in the "saving clause," it seems to me that we must look at Mr. Unlucky's present situation, and not at what happened in the past, to determine whether a § 2255 motion "is" currently "inadequate or ineffective." See Jennifer Case, Text Me: A Text-Based Interpretation of 28 U.S.C. § 2255(e) , 103 Ky. L. J. 169, 194 (2014–15) ("Another textual mistake that the circuit courts often make when interpreting § 2255(e)'s [s]aving [ ] [c]lause is to replace the verb 'is' with the word 'was.' ... When the linking verb is read (as Congress wrote it) in the present tense, the prisoner cannot access § 2241 unless § 2255 is—at the moment her § 2241 petition is filed in federal court—inadequate [or] ineffective to test the detention's legality.").

Contrary to the majority's approach, we must assess inadequacy and ineffectiveness as of the time Mr. Unlucky files his § 2241 habeas corpus petition, and not as of the time when he submitted his initial § 2255 motion. Otherwise, "Congress['] use of different sets of verbs, with distinct tenses ... would be pointless[.]" Freeman v. Quicken Loans, Inc. , 566 U.S. 624, 132 S.Ct. 2034, 2040, 182 L.Ed.2d 955 (2012). See also Carr v. United States , 560 U.S. 438, 448, 130 S.Ct. 2229, 176 L.Ed.2d 1152 (2010) ("Consistent with normal usage, we have frequently looked to Congress' choice of verb tense to ascertain a statute's temporal reach."); 1 U.S.C. § 1 ("In interpreting the meaning of any Act of Congress, unless the context indicates otherwise ... words used in the present tense include the future as well as the present.").

This temporal approach to the "saving clause" is not only linguistically proper, it is also historically sound. Before AEDPA, when res judicata did not bar successive § 2255 motions to vacate, see generally Salinger v. Loisel , 265 U.S. 224, 230–31, 44 S.Ct. 519, 68 L.Ed. 989 (1924), federal prisoners sometimes claimed that they could file a § 2241 habeas corpus petition because a motion to vacate would be "inadequate or ineffective" within the meaning of the last paragraph of § 2255 (which is identical to what is now § 2255(e) ). In response to such claims, at least some federal courts analyzed whether, at the time of the filing of the § 2241 petition, a § 2255 motion was "inadequate or ineffective." A good example is the discussion by the former Fifth Circuit in a case where the prisoner, having twice been denied § 2255 relief on a claim relating to his plea, filed a § 2241 habeas corpus petition. The former Fifth Circuit affirmed the dismissal of the § 2241 petition because the prisoner could file another § 2255 motion and have it considered by the sentencing court:

Neither of these [two § 2255 ] post conviction proceedings is subject to res judicata as such. Consequently, even though the petitioner may have presented this contention to the sentencing court on previous occasions, he is free to assert it again. We have no doubt that the District Court for the Western District of North Carolina will accord a full and fair hearing and, if appropriate, a right of appeal to the Fourth Circuit in the event that its decision on the merits is adverse to petitioner[.]

Birchfield v. United States , 296 F.2d 120, 122 (5th Cir. 1961) (emphasis added and citations omitted).¹ B

1 For other cases applying a similar temporal scope to the "saving clause," see, e.g., United States v. Hayman, 342 U.S. 205, 223, 72 S.Ct. 263, 96 L.Ed. 232 (1952) ("Nothing has been shown to warrant our holding at this stage of the proceeding that the [§] 2255 procedure will be 'inadequate or ineffective' if respondent is present for a hearing in the [d]istrict [c]ourt on remand of this case.") (emphasis added); Waugaman v. United States, 331 F.2d 189, 191 (5th Cir. 1964) (rejecting prisoner's argument that a § 2255 motion would be "inadequate or ineffective": "But if and when [the claims] are presented with sufficient factual particularity, we are confident that the Southern District of Ohio and if necessary on appeal, the Court of Appeals for the Sixth Circuit, would take cognizance of two things. ... There is, therefore, every assurance that the sentencing [c]ourt and the Sixth Circuit will accord the hearing the law requires and will grant the relief which the circumstances justify.") (emphasis added); and Johnson v. United States, 447 F.2d 516, 517 (5th Cir. 1971) (dismissing habeas corpus petition because prisoner had not sought to file a § 2255 motion to vacate: "Nevertheless, we note that Johnson has never had his case for post-conviction relief heard on the merits. Our decision is in no way intended to preclude Johnson from filing a properly designated § 2255 motion in the District Court for the District of Wyoming.") (emphasis added).

The majority says that the word "or" (in the phrase "inadequate or ineffective") merely introduces "ineffective" as a "synonym or 'definitional equivalent' " for "inadequate" (which the majority has replaced with "inadequate remedy at law"). See Maj. Op. at 1187–88. Like Judge Rosenbaum, see Rosenbaum Dissent at 1130, I disagree. The Supreme Court has told us that the "ordinary use [of the word 'or'] is almost always disjunctive, that is, the words it connects are to be given separate meanings." Loughrin v. United States , ––– U.S. ––––, 134 S.Ct. 2384, 2390, 189 L.Ed.2d 411 (2014) (rejecting an argument that would "construe ... two entirely distinct statutory phrases that the word 'or' joins as containing an identical element"). So, grammatically, words separated by "or" should not be read as duplicative of one another.

Even if the majority is correct that "inadequate" and "ineffective" are interchangeable, I'm not sure its textual analysis is correct. When a statutory term or phrase is undefined, courts try to ascertain its ordinary understanding at the time of enactment. See, e.g. , Perrin v. United States , 444 U.S. 37, 42, 100 S.Ct. 311, 62 L.Ed.2d 199 (1979). Because the majority looks to 1948—when the provision that is now § 2255(e) was first enacted—to determine the ordinary meaning of "inadequate" and "ineffective," I will do the same and consider what those words meant almost seven decades ago.

In the 1940s and 1950s, the word "inadequate" meant "[i]nsufficient; disproportionate; lacking in effectiveness or in conformity to a prescribed standard or measure." Black's Law Dictionary 902 (4th ed. 1951) (emphasis added). See also The Concise Oxford Dictionary of Current English 573 (3d ed. 1944) ("Not adequate (to purpose); insufficient."); Webster's New Collegiate Dictionary 419 (2d ed. 1949) ("Not adequate; deficient; insufficient."). So, contrary to the majority's suggestion, one definition of "inadequate" is tied to the ultimate effect, i.e., the result.

The word "ineffective," like the word "inadequate," was also not limited to process in the 1940s and 1950s. Indeed, some definitions linked the word with results. For example, one dictionary defined "ineffective" as "[n]ot producing the desired effect [.]" The Concise Oxford Dictionary of Current English 583 (3d ed. 1944) (emphasis added). According to another dictionary, "ineffective" meant "[n]ot effective; productive of no effect ; ineffectual[.]" Webster's New Collegiate Dictionary 428 (2d ed. 1949) (emphasis added).²

2 Insofar as "to test" is concerned, one dictionary in the 1940s and 1950s defined "test" as to "[p]ut to the test, make trial of[.]" The Concise Oxford Dictionary of Current English 1266 (3d ed. 1944). Another used a similar definition: "[t]o put to the test or proof; to try." Webster's New Collegiate Dictionary 878 (2d ed. 1949).

So, even if we assume that the words "inadequate" and "ineffective" meant (and mean) the same thing, that assumption does not help the majority. There is a strong textual argument that the phrase "inadequate or ineffective" is concerned with both procedure (i.e., process) and substance (i.e., results). The phrase can easily be read to have some relationship, some connection, to the ability of a § 2255 movant to file a motion to vacate, as well as his ability to obtain a desired substantive result.

C

Returning to our fictional case, Mr. Unlucky is unable to file a second or successive § 2255 motion to vacate based on the Supreme Court's interpretation of § 999.99. That is because § 2255(h) only permits a second or successive motion to vacate when there is newly discovered evidence demonstrating, by clear and convincing evidence, that no reasonable factfinder would have found the person guilty of the offense, or where there is a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable. As noted earlier, Mr. Unlucky does not have newly discovered evidence, and his claim is not based on a new rule of constitutional law.

If Mr. Unlucky, who is innocent of the charge that has landed him in prison, cannot even file a second or successive § 2255 motion—and by definition cannot succeed on such a motion—then "the remedy by motion" is presently "inadequate or ineffective" to test the legality of his detention. See In re Davenport , 147 F.3d 605, 611 (7th Cir. 1998) ("A procedure for postconviction relief can fairly be termed inadequate when it is so configured as to deny a convicted defendant any opportunity for judicial rectification of so fundamental a defect in his conviction as having been imprisoned for a nonexistent offense. It could indeed ... be thought an inadequacy of constitutional dimensions."); In re Dorsainvil , 119 F.3d 245, 251 (3d Cir. 1997) (recognizing that the "saving clause" may be broad enough to allow a § 2241 habeas corpus petition where "a defendant [is] imprisoned for a crime that an intervening decision [later] negates"). As I see it, Mr. Unlucky has satisfied the "saving clause," and can file a habeas corpus petition pursuant to § 2241 which relies on the new Supreme Court decision interpreting § 999.99. Cf. United States ex rel. Leguilllou v. Davis , 212 F.2d 681, 684 (3d Cir. 1954) (interpreting the last paragraph of the former version of § 2255 : "[W]e think the remedy by motion can be 'inadequate or ineffective to test the legality of ... detention' only if it can be shown that some limitation of scope or procedure would prevent a [§] 2255 proceeding from affording the prisoner a full hearing and adjudication of his claim of wrongful detention.").³

3 There are some other pre-AEDPA decisions analyzing what makes a § 2255 remedy "inadequate or ineffective," but they are not of much help here because in those cases there was a court available to consider the prisoner's motion to vacate. See, e.g., Adam v. Hagan, 325 F.2d 719, 720 (5th Cir. 1963) (holding that the distance between the place of confinement and the sentencing court does not make a § 2255 motion "inadequate or ineffective"); Scott v. Welch, 192 F.2d 676, 677 (4th Cir. 1951) (concluding that the denial of IFP status does not render a § 2255 motion "inadequate or ineffective"). Cf. Swain v. Pressley, 430 U.S. 372, 381–83, 97 S.Ct. 1224, 51 L.Ed.2d 411 (1977) (interpreting a similar "saving clause" in a provision of the D.C. Code and ruling that a post-conviction remedy is not "inadequate or ineffective" just because it is resolved by an Article I court).

D

In order to avoid constitutional problems, the "saving clause" of § 2255(e) must have some meaning. See Boumediene v. Bush , 553 U.S. 723, 776, 128 S.Ct. 2229, 171 L.Ed.2d 41 (2008) ; Swain , 430 U.S. at 381–82, 97 S.Ct. 1224 ; United States v. Hayman , 342 U.S. 205, 223, 72 S.Ct. 263, 96 L.Ed. 232 (1952). We have a duty to "give effect, if possible, to every clause and word of a statute," Duncan v. Walker , 533 U.S. 167, 174, 121 S.Ct. 2120, 150 L.Ed.2d 251 (2001), and if the "saving clause" did not amount to anything in a post-AEDPA world, Congress likely would not have carried it over wholesale in 1996. What makes this case hard is figuring out the proper interplay (and balance) between the "saving clause" of § 2255(e) and the restrictions that § 2255(h) places on second or successive motions to vacate. The content of the "saving clause" must be meaningful, but not so broad that it swallows § 2255(h).

My solution is to draw upon the undisputed—but too often forgotten—principle that "habeas corpus is, at its core, an equitable remedy," Schlup v. Delo , 513 U.S. 298, 319, 115 S.Ct. 851, 130 L.Ed.2d 808 (1995), and read the "saving clause" to allow an innocent person like Mr. Unlucky to obtain § 2241 habeas relief. First, from the earliest days of the Republic, the Supreme Court has granted habeas corpus relief and ordered the discharge of federal prisoners where the facts alleged by the government did not constitute a federal crime. See Ex parte Bollman , 8 U.S. (4 Cranch) 75, 136, 2 L.Ed. 554 (1807) (Marshall, C.J.) ("[A]s the crime [of treason] with which the prisoners stand charged has not been committed, the court can only direct them to be discharged."). Second, the importance of innocence runs deep in our habeas jurisprudence, and there is nothing more inequitable than having a person serve a sentence in a federal prison for a non-existent crime. That is why the Supreme Court, pre-AEDPA, ruled that claims of innocence based on new statutory interpretations can be asserted under § 2255. See Davis v. United States , 417 U.S. 333, 346, 94 S.Ct. 2298, 41 L.Ed.2d 109 (1974) ("If this contention [about the intervening change in law] is well taken, then Davis' conviction and punishment are for an act that the law does not make criminal. There can be no room for doubt that such a circumstance inherently results in a complete miscarriage of justice and present[s] exceptional circumstances that justify collateral relief under § 2255.") (internal quotation marks omitted).

Even those who advocated for a narrower scope of federal habeas review prior to AEDPA recognized that the writ should be available in cases where "a convicted defendant makes a colorable showing that an error, whether constitutional or not, may be producing the continued punishment of an innocent man." Henry Friendly, Is Innocence Irrelevant?: Collateral Attacks on Criminal Judgments , 38 U. Chi. L. Rev. 142, 160 (1970). I cannot believe that a federal court, with the "saving clause" available, would deny habeas corpus relief to a person who does not have a § 2255 remedy available but can show, based on a new (and governing) decision, that he never committed a federal crime.⁴

4 Innocence is so strong a concept that, when sufficiently proven, it even constitutes a judge-made vehicle for avoiding hurdles like procedural default and untimeliness. See, e.g., McQuiggin v. Perkins, ––– U.S. ––––, 133 S.Ct. 1924, 1931–35, 185 L.Ed.2d 1019 (2013) ; House v. Bell, 547 U.S. 518, 522, 126 S.Ct. 2064, 165 L.Ed.2d 1 (2006).

Reading the "saving clause" to allow only claims of statutory innocence under § 2241 does relatively little harm to the structure of § 2255. It gives the "saving clause" a narrow but important scope, and does not do too much violence to § 2255(h)'s restrictions on second or successive motions to vacate.

The line I have drawn, admittedly, is not perfect. If Mr. Unlucky is able to seek habeas corpus relief because a new (and governing) statutory ruling shows that he never committed a crime, why shouldn't the writ also be available—as Judges Martin and Rosenbaum contend—when a new statutory decision by the Supreme Court makes it clear that a defendant's sentence exceeds the statutory maximum? I confess that I don't have very good answers to that question, but equity—with its concern for justice—does not always draw clean lines, and the finality concerns embodied in § 2255(h) cannot be ignored. If we are going to allow any federal prisoners to use the habeas remedy pursuant to the "saving clause" of § 2255(e), it should be those who are languishing in prison despite having never committed a crime. A criminal justice system run by fallible human beings can tolerate most non-capital sentencing errors, but it cannot, I submit, refuse to hear the claims of those incarcerated for non-existent offenses.

III

The majority, understanding that § 2255(e)'s "saving clause" must allow for habeas corpus relief in some circumstances to avoid being illusory, carves out some territory where it posits that the "saving clause" can override § 2255(h)'s restrictions. It offers two main examples of scenarios where § 2241 can be used. First, the majority says that § 2255(e) may be used by federal prisoners challenging determinations about parole or good-time credits. See Maj. Op. at 1089, 1092–93. Second, the majority, like the Tenth Circuit in Prost v. Anderson , 636 F.3d 578, 588 (10th Cir. 2011), says that § 2255(e) allows a federal prisoner to file a habeas petition when his or her sentencing court is no longer available, such as when a military prisoner's tribunal has been dissolved. See id. at 1092–93. The problem is that these examples ignore the "authorization clause" of § 2255(e).

As explained in Andrews v. United States , 373 U.S. 334, 338, 83 S.Ct. 1236, 10 L.Ed.2d 383 (1963), the "motion" that § 2255(e) refers to is a motion to vacate filed by a federal prisoner in the federal court that imposed the sentence. See also Yirkovsky v. Gonzales , 2007 WL 2476766, at *1 (D.S.D. Aug. 27, 2007) ("Petitioner is 'authorized to apply for relief by motion' pursuant to [ ] § 2255 because he is a prisoner in custody pursuant to a federal conviction and sentence who may move the [sentencing] court that imposed the sentence to vacate, set aside or correct the sentence[.]"). As a result, § 2255(e)"operates to bar a § 2241 habeas petition only if § 2255 authorizes the prisoner to bring a ... motion [to vacate]. Importantly, if the [a]uthorization [c]lause is not satisfied, subsection (e) plays no role in determining whether a prisoner can bring his habeas petition." Case, Text Me , 103 Ky. L. J. at 187.⁵

5 Judge Rosenbaum aptly explains in her dissenting opinion that not all federal collateral claims are "authorized" by motion under § 2255. See Rosenbaum Dissent at 1126–28. And, significantly, not all federal prisoners are authorized to file a § 2255 motion. Instead, "[o]nly federal prisoners who have been 'sentence[d]' by a federal court are eligible." See id. at 1126.

The majority incorrectly assumes that prisoners challenging determinations about parole and good-time credits, or attacking a sentence imposed by a military tribunal that no longer exists, can file a motion to vacate under § 2255. In the words of § 2255(e)'s "authorization clause," prisoners in these two scenarios were never "authorized to apply for relief by motion pursuant to this section [i.e., § 2255]" in the first place. And I'm not sure that prisoners sentenced in dissolved territorial courts are any different. A statutory "saving clause" (like the one in § 2255(e) ) is a carve-out from the general requirements of a statute, and if the statute does not apply to begin with, then the "saving clause" never comes into play.

A

Federal prisoners challenging determinations about parole and good-time credits can seek habeas corpus relief pursuant to § 2241. See, e.g. , Granville v. Hogan , 591 F.2d 323, 324 (5th Cir. 1979) (good-time credits); Gomori v. Arnold , 533 F.2d 871, 874–75 (3d Cir. 1976) (calculation of release date); Zannino v. Arnold , 531 F.2d 687, 690–91 (3d Cir. 1976) (parole); Halprin v. United States , 295 F.2d 458, 459 (9th Cir. 1961) (parole). But that does not mean that those prisoners can do so because of § 2255(e).

In fact, prisoners challenging determinations about parole or good-time credits have always had to proceed under § 2241 and have never been able to file motions to vacate under § 2255. See, e.g. , Hajduk v. United States , 764 F.2d 795, 796 (11th Cir. 1985) ("A challenge to the lawfulness of the parole commission's actions cannot be brought pursuant to 28 U.S.C. § 2255. Hajduk's ex post facto argument is nothing more than a challenge to the lawfulness of the parole commission's actions, not the lawfulness of the sentence imposed by the court. Such an action must be brought as a petition for writ of habeas corpus pursuant to 28 U.S.C. § 2241."). As a result, Judge Rosenbaum is correct that such prisoners do not come within the "authorization clause" of § 2255(e), and therefore do not need the "saving clause" to avail themselves of a habeas remedy. See Rosenbaum Dissent at 1127–28.

B

The same is true of prisoners challenging a conviction secured in a military tribunal. Like federal prisoners who wish to challenge determinations about parole and good-time credits, federal prisoners convicted and sentenced in military tribunals have long been able to file traditional habeas corpus petitions. See, e.g. , United States v. Augenblick , 393 U.S. 348, 350, 89 S.Ct. 528, 21 L.Ed.2d 537 (1969) ; Burns v. Wilson , 346 U.S. 137, 139–142, 73 S.Ct. 1045, 97 L.Ed. 1508 (1953) ; Gusik v. Schilder , 340 U.S. 128, 129, 71 S.Ct. 149, 95 L.Ed. 146 (1950) ; Carter v. McClaughry , 183 U.S. 365, 366–67, 22 S.Ct. 181, 46 L.Ed. 236 (1902) ; Ex parte Milligan , 71 U.S. 4 Wall. 2, 68–69, 18 L.Ed. 281 (1866) ; Calley v. Callaway , 519 F.2d 184, 199 (5th Cir. 1975) (en banc).

Allowing a federal military prisoner to file a petition for writ of habeas corpus makes sense because a court-martial (or similar military tribunal) "is a special body convened for a specific purpose, and when that purpose is accomplished its duties are concluded and the court is dissolved." McClaughry v. Deming , 186 U.S. 49, 64, 22 S.Ct. 786, 46 L.Ed. 1049 (1902). Access to habeas for such prisoners, however, does not come from (or run into the limitations of) § 2255, which is reserved for federal prisoners convicted in, and sentenced by, federal courts.

As noted, the majority cites the Tenth Circuit's decision in Prost with approval, but in my opinion this aspect of Prost is flawed. Prost relied on an earlier Tenth Circuit decision, Ackerman v. Novak , 483 F.3d 647 (10th Cir. 2007), as support for its suggestion that a military prisoner may resort to the "saving clause" and file a § 2241 petition where a § 2255 motion "ha[s] to be brought in the (now nonexistent) sentencing court, [and] that remedial mechanism [is] necessarily inadequate and ineffective to test the legality of his detention...." Prost , 636 F.3d at 588. But the panel in Prost missed the holding of Ackerman. In Ackerman , a federal prisoner convicted by a military court-martial sought authorization from the Tenth Circuit, see 28 U.S.C. § 2244(a), to file a second or successive habeas corpus petition under 28 U.S.C. § 2254. See Ackerman , 483 F.3d at 648–49. The Ackerman panel first explained, in no uncertain terms, that a prisoner convicted in, and sentenced by, a military tribunal can seek collateral review only by way of a habeas corpus petition under § 2241. See id. at 649. Such a prisoner cannot use § 2254 because that provision is reserved for prisoners in state custody, and cannot use § 2255 because his military tribunal has dissolved and cannot entertain a collateral attack. See id. at 649–50 & n.2. The Ackerman panel concluded that, because a military court-martial is not a "court of the United States" within the meaning of § 2244(a), the prisoner did not need to obtain circuit authorization to file a § 2241 habeas corpus petition. See id. at 651–53.

Ackerman , then, provides no support for the claim by Prost , and by the majority here, that a military prisoner needs the "saving clause" of § 2255(e) to file a § 2241 habeas corpus petition. Simply stated, a military prisoner has a § 2241 remedy that is available independent of § 2255. See Clinton v. Goldsmith , 526 U.S. 529, 537 n.11, 119 S.Ct. 1538, 143 L.Ed.2d 720 (1999) ("[O]nce a criminal conviction has been finally reviewed within the military system, and a servicemember in custody has exhausted other avenues provided under the [Code of Military Justice] to seek relief from his conviction, he is entitled to bring a habeas corpus petition, see 28 U.S.C. § 2241(c), claiming that his conviction is affected by a fundamental defect that requires that it be set aside.") (citations omitted); Palomera v. Taylor , 344 F.2d 937, 938 (10th Cir. 1965) ("A motion under 28 U.S.C. § 2255 is not proper here because the petitioner was sentenced by a military court-martial convened in 1944.").

C

As for prisoners convicted in territorial courts that no longer exist, that is a more nuanced matter. But it is not clear to me that such prisoners need the "saving clause" to file a § 2241 habeas corpus petition. First, if the territorial courts in question are created by the legislature of the territory, then they are not "courts established by an Act of Congress" within the meaning of § 2255(a). See In re Moran , 203 U.S. 96, 104, 27 S.Ct. 25, 51 L.Ed. 105 (1906) (Territory of Oklahoma); Connella v. Haskell , 158 F. 285, 287 (8th Cir. 1907) (same). Such courts, therefore, could never entertain a § 2255 motion in the first place, meaning that the "authorization clause" of § 2255(e) would not be satisfied. Second, if territorial jurisdiction is being exercised by federal district courts located in the territory pursuant to an act of Congress, then those district courts can entertain a § 2255 motion to vacate, and the scenario suggested by the majority—that of a sentencing court that no longer exists—is more imagined than real. See Madsen v. Hinshaw , 237 F.2d 370, 371 (9th Cir. 1956) (Territory of Alaska).

In the relatively unusual scenario where a territory becomes a state, and the federal courts in the new state refuse to entertain § 2255 motions by prisoners previously convicted of territorial crimes, a motion to vacate may be "inadequate or ineffective" within the meaning of the "saving clause." See, e.g. , Spaulding v. Taylor , 336 F.2d 192, 193 (10th Cir. 1964) (federal district court in Alaska, following admission to statehood, refused to consider motion to vacate, thereby allowing prisoner convicted of territorial crime to seek a writ of habeas corpus pursuant to § 2241 ). Even in such circumstances, however, there are more questions than answers. For example, in a case very similar to Spaulding , the former Fifth Circuit explained that the state courts of Alaska, following Alaska's admission to statehood, were willing to consider post-conviction motions filed by prisoners previously convicted of territorial offenses. See Hutson v. Zeigler , 362 F.2d 200, 204 & n.9 (5th Cir. 1966). As a result, those prisoners had to proceed under § 2254 in federal court but first had to exhaust their claims in the Alaska state courts. See id. at 204.

Given this tapestry, I do not understand what possible application the "saving clause" has under the majority's rationale. It seems to me that the majority has come dangerously close to sapping the "saving clause" of any meaning.

IV

I read the "saving clause" of § 2255(e) to permit § 2241 habeas corpus petitions by federal prisoners who can no longer file a motion to vacate and who, based on a new (and governing) statutory decision, are in custody despite never having committed a crime. Because Mr. McCarthan is not asserting such a claim of innocence, I concur in the judgment.

WILSON, Circuit Judge, joined by JILL PRYOR, Circuit Judge, dissenting:

I would reverse the denial of Dan McCarthan's claim and remand for the district court to consider the merits in the first instance. I am, for the most part, persuaded by Judge Jordan's interpretation of the savings clause. I agree with his textual analysis of the clause, but I believe the equitable nature of the Great Writ dictates a different result than he reaches.

Judge Jordan states that the equitable nature of the Writ leads him to conclude that the savings clause applies to a prisoner who asserts a claim of actual innocence but not to a prisoner who, like McCarthan, argues that his sentence exceeds the statutory maximum. In my opinion, the savings clause applies to both types of prisoners. A prisoner who is actually innocent is in the same position as a prisoner whose sentence exceeds the statutory maximum—each prisoner is being deprived of his liberty even though no law authorizes the deprivation. As my colleague Judge Hill once said: "If a petitioner can show that he is illegally incarcerated, he is entitled to release. Fairness requires it. Justice is the ultimate goal in the grant of the Writ." Rozier v. United States , 701 F.3d 681, 690 (11th Cir. 2012) (Hill, J., dissenting).

Justice demands that, at the very least, McCarthan receive a chance to "test the legality of his detention." See 28 U.S.C. § 2255(e).

MARTIN, Circuit Judge, joined by JILL PRYOR, Circuit Judge, dissenting.

Dan McCarthan was sentenced to serve 211 months (17.5 years) in prison based on this Court's mistake of the law. Mr. McCarthan had been convicted of felony offenses earlier in his life, and he was found with a firearm, so he was due to go to prison. Ordinarily, a felon convicted of possessing a firearm faces up to 10 years in prison, but no more. 18 U.S.C. § 924(a)(2). On the other hand, when the person has three earlier convictions for crimes that are either a "violent felony" or a "serious drug offense," the Armed Career Criminal Act (ACCA), id. § 924(e), increases his sentence to no less than 15 years and up to life. Because Mr. McCarthan was sentenced under ACCA as though he had three qualifying convictions, he got a sentence that was seven and one-half years longer than the statute would have otherwise allowed. He got this much longer sentence because in his past he had been convicted of walkaway escape,^{1 1} and Eleventh Circuit precedent characterized walkaway escape as a "violent felony." See United States v. Gay , 251 F.3d 950, 953–55 (11th Cir. 2001) (per curiam). But this Court's ruling in Gay was wrong. Well after Mr. McCarthan began serving his 17.5-year sentence, the Supreme Court taught us that escape is not a "violent felony." Its decisions in Begay v. United States , 553 U.S. 137, 128 S.Ct. 1581, 170 L.Ed.2d 490 (2008), and Chambers v. United States , 555 U.S. 122, 129 S.Ct. 687, 172 L.Ed.2d 484 (2009), overturned our decision in Gay . See United States v. Lee , 586 F.3d 859, 874–75 (11th Cir. 2009).

1 In 1992, Mr. McCarthan was convicted of escape in Florida for walking away from an unsecured correctional facility without permission. The PSR describes his escape conviction as follows:

According to court records, on February 14, 1988, the defendant signed out for work from the Tampa Community Corrections Center with a return time of 1:30 a.m. on February 15, 1998 [sic]. He failed to return to [sic] by 1:30 a.m., as required. The defendant returned to the center at 12:58 p.m. on February 15, 1998 [sic]. The escape report was canceled. At 3:30 p.m. on February 15, 1988, the defendant left the center without permission, and an escape report was again initiated.

1 Our Circuit has, in the past, referred to the clause as the "savings clause." See, e.g., Wofford, 177 F.3d at 1237. I agree with the Majority, see Maj. Op. at 1081–82, that we should refer to it as the "saving clause." Indeed, the Supreme Court has called this clause the "saving clause." Boumediene v. Bush, 553 U.S. 723, 776, 128 S.Ct. 2229, 171 L.Ed.2d 41 (2008).

Mr. McCarthan is now asking us to make right what we caused to go wrong when he got his 17.5-year sentence. The panel that first heard Mr. McCarthan's case (I was a member) applied Eleventh Circuit precedent. See McCarthan v. Warden, FCI Estill , 811 F.3d 1237 (11th Cir. 2016). Under that precedent, the panel concluded that he could not satisfy the jurisdictional test our Court created for habeas cases in which a prisoner is seeking relief from a wrongly imposed ACCA sentence. So we denied him relief. Id. at 1256–57. The government did not seek rehearing from our ruling. Indeed, the United States subscribed to this Circuit's habeas jurisprudence as set out in the McCarthan panel opinion, and anyway it had won.

Nevertheless, a majority of this Court voted to vacate the panel's opinion and hear Mr. McCarthan's case en banc. Since the government never asked us to rehear his case, one might think en banc rehearing would be good news for Mr. McCarthan. But today's majority opinion not only does nothing to change Mr. McCarthan's loss into a win, it puts relief out of reach for others who have been sentenced based on a legal mistake. In other words, this Court voted to rehear Mr. McCarthan's case not because the Court believed the panel was wrong in its application of our circuit precedent, but instead because the Majority wanted to overturn that precedent. Before today, the path to relief for prisoners like Mr. McCarthan has been narrow, indeed. Today's majority opinion cuts off that path entirely.

The Majority concludes that Mr. McCarthan's claim cannot be recognized under the federal statute that governs postconviction challenges by federal prisoners. My colleagues in dissent say that the law recognizes his claim, and would remand Mr. McCarthan's case to be evaluated anew on the merits. My view is slightly different, so I write separately to say how I believe Mr. McCarthan's case should turn out, and why.

I. THE SAVINGS CLAUSE

The majority opinion characterizes this case as a rather dry and complex exercise in statutory construction. A reader could almost miss the fact that what we are talking about is who, among the hundreds of thousands of human beings incarcerated in U.S. prisons, will have access to relief under a writ of habeas corpus.^{2 2} The writ of habeas corpus is of such fundamental importance to this nation's legal system that it is known as the Great Writ. See Ex parte Bollman , 8 U.S. (4 Cranch) 75, 95, 2 L.Ed. 554 (1807) (Marshall, C.J.). The writers of our Constitution recognized the importance of the writ of habeas corpus when they enshrined its existence in that document. U.S. Const. art. I, § 9, cl. 2. The Great Writ is the tool meant to be available to any person who finds himself in jail when he ought not be there.

2 The Department of Justice estimates that in 2015 (the latest year for which it has published statistics), there were 328,500 people in federal correctional custody. See Bureau of Justice Statistics, U.S. Dep't of Justice, Correctional Populations in the United States, 2015 , at 12 (Dec. 2016).

2 My analysis does not foreclose the possibility that a constitutional deficiency of another type may allow for other kinds of second or successive claims to be considered under the saving clause. But because this case involves only a second or successive claim that relies on a retroactively applicable new rule of law, I do not analyze what types of other claims, if any, might so qualify.

In 1948, for reasons explained by the Majority and Judge Rosenbaum, 28 U.S.C. § 2255 was enacted. Then in 1996, the Antiterrorism and Effective Death Penalty Act (AEDPA) added limitations to § 2255 which remain in effect today. Under 28 U.S.C. § 2255, once a federal prosecution results in a final conviction, the prisoner is generally allowed to challenge the legality of his detention only through a § 2255 motion, and not through a petition for writ of habeas corpus. See 28 U.S.C. § 2255(e) ("An application for a writ of habeas corpus in behalf of a prisoner who is authorized to apply for relief by motion pursuant to this section, shall not be entertained...."). Significant to many cases like Mr. McCarthan's is that the relief offered by § 2255 is narrowly defined and tightly administered. A prisoner is generally allowed to challenge his conviction and sentence by way of a § 2255 motion just one time, id. § 2255(h), and that challenge generally must be made within one year of his conviction becoming final, id. § 2255(f). While the statute's goals of tight deadlines and finality might seem desirable, they were implemented at the same time the federal prison population was exploding; federal sentences were getting longer; and the U.S. Sentencing Guidelines were requiring judges to make many rulings before arriving at the sentence imposed.^{3 3} So it is a fact of life for these prisoners that they can sit in jail for years or even decades before the Supreme Court comes to tell inferior federal courts (like this one) about a mistake the court made when a sentence was imposed. This has happened to Mr. McCarthan and so many others. This Court was wrong when it said that Mr. McCarthan's earlier conviction for walkaway escape required his sentence to be (significantly) longer than the ten-year cap called for by the statute that otherwise would have governed his sentence. So while this case is about how we construe the words of a statute, it is also about whether Mr. McCarthan and those like him should continue to bear the burden of the mistake the federal courts made in sentencing him. I part ways with the Majority, because I think not. And while the Majority highlights the rule in the Tenth Circuit, the fact is that most every other U.S. Court of Appeals to have reached this question thinks not as well.

3 Sections 2255(f) and (h) were passed in 1996 as part of AEDPA. In 1990, there were 58,838 people incarcerated in federal prisons; a decade later, that number had risen to 133,921. See Bureau of Justice Statistics, U.S. Dep't of Justice, Prisoners in 2000 2 (Aug. 2001). As mentioned above, the latest available data show that there are now approximately 328,500 people in federal prison. Supra note 2.

3 This must be a § 2255 motion, as opposed to any other kind of motion, for two reasons: (1) the preceding term "by motion" is shorthand for the "by motion pursuant to this section" phrase used earlier in § 2255, and (2) no other type of relevant motion would be filed with the sentencing court.

The History of Section 2255 in the Eleventh Circuit

Section 2255 generally allows a prisoner to bring a new attack (the statute uses the term "second or successive motion") on his conviction only if his claim falls into one of the two narrow categories in § 2255(h). That is: (1) a claim of actual innocence based on newly discovered evidence, id. § 2255(h)(1); or (2) a claim based on "a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable," id. § 2255(h)(2).

Prisoners like Mr. McCarthan, who want to challenge their detention on the basis of a new, retroactive statutory interpretation by the Supreme Court, do not fall under either category of § 2255(h). He is hanging his hat on a new rule of statutory not constitutional law. Since he cannot proceed under § 2255(h), he seeks to proceed under the "savings clause"^{4 4} of § 2255(e). The savings clause says a prisoner can bypass the constraints of § 2255(h) and file a habeas petition challenging his detention if it "appears that the remedy by [§ 2255 ] motion is inadequate or ineffective to test the legality of his detention." Id. § 2255(e). That leaves us to decide for Mr. McCarthan (and many others): when is § 2255"inadequate or ineffective to test the legality of [a prisoner's] detention"?

4 My colleagues who I join in writing about Mr. McCarthan's case have adopted the term "saving clause" as opposed to the term this Court has always used: "savings clause." See, e.g. , Mackey v. Warden, FCC Coleman-Medium , 739 F.3d 657, 661 (11th Cir. 2014) ("[The] exception to § 2255(e)'s bar on a § 2241 petition is commonly referred to as the 'savings clause.' "); Samak v. Warden, FCC Coleman-Medium , 766 F.3d 1271, 1279 (11th Cir. 2014) (Pryor, William, J., concurring) ("That exception—the 'savings clause'—recognizes that a motion to vacate a sentence may sometimes be inappropriate, that is, 'inadequate or ineffective,' so in that circumstance Congress allows a federal prisoner to 'test the legality of his detention' in the traditional action against his custodian."); Bryant v. Warden, FCC Coleman-Medium , 738 F.3d 1253, 1262 (11th Cir. 2013) ("[The] exception to § 2255(e)'s bar on a § 2241 petition is commonly referred to as the 'savings clause.' "); Gilbert v. United States , 640 F.3d 1293, 1305–06 (11th Cir. 2011) (en banc) ("The [ ] exception to the § 2255(e) bar on § 2241 petitions, commonly referred to as the 'savings clause,' is the focus of our issue[.]").
While I recognize that the Supreme Court used the term "saving clause" in Boumediene , courts still use both "saving" and "savings" in this context. I will continue as we have, not because of any value judgment about the Majority's new term, but because I have grown accustomed to the old.

4 Of course, the language of the saving clause also anticipates the filing of some initial claims. The statute employs the language, "the applicant has failed to apply for relief, by motion, to the court which sentenced him," which means that the applicants to whom it refers are first-time filers in the sense that they have not yet filed a § 2255(a) motion with the court that sentenced them. This language appears in the disjunctive, as an alternative to "such court has denied him relief." So first-time claimants, like applicants who bring second or successive claims, also are entitled to bring a § 2241 petition under the saving clause, provided the saving clause is otherwise satisfied. These first-time claimants would be those who, for practical reasons, cannot obtain "adequate" relief through a § 2255(a) proceeding in the court that sentenced them. See infra at 1130–32.

A lot of ink has spilled and many lives have been touched as a result of this Court's work on how to apply the savings clause. The Court offered its first interpretation of the savings clause in Wofford v. Scott , 177 F.3d 1236 (11th Cir. 1999). Wofford established that, in order for a prisoner to rely on the savings clause, he had to show his claim had been "squarely foreclosed" by circuit law at the time of his trial, appeal, and first § 2255 motion. 177 F.3d at 1244. In other words, the prisoner was required to show that the courts in this Circuit would have ruled against him on this claim at the time he was convicted and sentenced, and when he appealed, and when he filed for postconviction relief by way of a § 2255 motion. To the extent I have been involved in these cases, I have always believed that Wofford was wrong and that this Court's rulings on savings clause cases that have since followed Wofford are wrong as well. See Bryant , 738 F.3d at 1300 (Martin, J., concurring in part and dissenting in part); Williams v. Warden, Fed. Bureau of Prisons , 713 F.3d 1332, 1350–56 (11th Cir. 2013) (Martin, J., dissenting); Gilbert , 640 F.3d at 1330–36 (Martin, J., dissenting).

Wofford 's"squarely foreclosed" requirement became the bedrock of this Court's savings clause jurisprudence. See Bryant , 738 F.3d at 1272 ("What makes the § 2255 proceeding 'inadequate or ineffective' for petitioner Bryant is that he had no 'genuine opportunity' to bring his § 924(e) claim because Circuit precedent squarely foreclosed that claim throughout his trial, direct appeal, and first § 2255 motion.").^{5 5} Under this Court's "squarely foreclosed" requirement, each time we consider a prisoner's claim for savings clause relief, we must look backward and ask whether the petitioner's original § 2255 proceeding was "inadequate or ineffective to test the legality of his detention." But the savings clause nowhere requires us to do this. Rather, the savings clause says that the writ of habeas corpus is preserved for cases in which § 2255"is inadequate or ineffective to test the legality of [the prisoner's] detention." 28 U.S.C. § 2255(e). The plain text of the statute allows a prisoner to seek habeas relief when § 2255 is "inadequate or ineffective" to bring his current challenge to the legality of his detention. When the statute is read as Congress wrote it, in the present tense, it is clear that a prisoner can bring a habeas petition if § 2255 is—at the time the petition is filed in federal court—"inadequate or ineffective" to test the legality of the detention.

5 In Bryant , this Court created a five-part test a prisoner must pass before he is allowed to access the savings clause to make a claim, like the one Mr. McCarthan makes here: that one of his previous convictions was wrongly characterized as a "violent felony" under § 924(e) causing him to receive a sentence of at least 15 years under § 924(e) rather than a sentence of no more than 10 years under § 924(a). This five-part test is found nowhere in the words of the statute. It was this five-part Bryant test that required the panel to deny relief to Mr. McCarthan.
The first step of the Bryant test is the "squarely foreclosed" requirement. See Bryant , 738 F.3d at 1274 ("[The petitioner] must establish that ... throughout his sentencing, direct appeal, and first § 2255 proceeding, our Circuit's binding precedent had specifically addressed [his] distinct prior state conviction that triggered § 924(e) and had squarely foreclosed [his] § 924(e) claim that he was erroneously sentenced above the 10–year statutory maximum penalty in § 924(a)."). The second step of the Bryant test further enforces the "squarely foreclosed" requirement. See id. ("[The petitioner] must establish that ... subsequent to his first § 2255 proceeding, [a] Supreme Court [ ] decision ..., as extended by this Court to [his] distinct prior conviction, overturned our Circuit precedent that had squarely foreclosed [his] § 924(e) claim.").

5 After today's decision, obviously, at least some aspects of Williams 's interpretation of § 2255(e) are no longer valid. Whether the determination that § 2255(e) is jurisdictional withstands our sua sponte abrogation of our prior interpretation of § 2255(e) in cases such as Williams is unclear.

The Majority is right when it says that the rules this Court created for these cases have not worked well. I have seen the problems resulting from this circuit's "squarely foreclosed" rule play out over the years. For example, in Albert Williams's 2013 appeal, this Court left him to serve a 293-month sentence (more than 24 years), rather than a sentence that, by law, should have been capped at 10 years. Williams , 713 F.3d at 1334. Relief had to be denied to him, this Court said, because at the time he challenged his sentence on direct appeal, and then again at the time he filed his § 2255 motion, our Court had never decided the issue of whether his prior convictions for burglary should be considered "violent felonies" to enhance his sentence. Id. at 1348. The Williams panel blinded itself to what the state of the law was at the time we ruled on his § 2255 motion in 2013. We said that because no Eleventh Circuit precedent had ruled on whether a Florida burglary conviction is an ACCA-qualifying offense at the time of his direct appeal in 1999, his claim was not "squarely foreclosed." Id. That meant, so the logic went, that a § 2255 motion would not have been ineffective as a way to raise the claim, and so Mr. Williams was not entitled to relief. Id . at 1345. For my part as a member of the Williams panel, I asked how in the world this Court's lack of having ruled on a question in the past could possibly give us the power to keep Mr. Williams in prison for more than 24 years when Congress never gave us the power to keep him in prison for more than 10. Id. at 1353 (Martin, J., dissenting) ("The correct question to ask is whether Mr. Williams was erroneously sentenced as an armed career criminal in light of Begay . If he was, the federal courts never had jurisdiction to sentence him above the 10 year maximum allowed by law. The existence or nonexistence of circuit precedent which conflicts with Begay cannot operate to confer jurisdiction on this Court."). Now years have passed and this Court has only recently resolved the question of whether Florida burglary (the same statute Mr. Williams was litigating back in 2013) is a "violent felony" for purposes of ACCA, and held it is not. See United States v. Esprit , 841 F.3d 1235, 1237 (11th Cir. 2016). Meanwhile, Mr. Williams remains in prison based on this mistake we made in lengthening his sentence.⁶

6 Mr. Williams's quest for relief on this issue continues. Since the Supreme Court gave retroactive relief to some inmates serving sentences improperly enhanced under ACCA, this Court granted Mr. Williams permission to file a second or successive petition on July 1, 2016.

I read § 2255 to allow a prisoner to file a habeas petition under the savings clause when he shows that, at some point after his first § 2255 proceeding, there was a retroactive decision from an authoritative federal court, which interpreted a statute in a way that now reveals a fundamental defect in that prisoner's conviction or sentence. Both the Sixth and Seventh Circuits interpret the savings clause this way. See Hill v. Masters , 836 F.3d 591, 595 (6th Cir. 2016) ("When seeking to petition under § 2241 based on a misapplied sentence, the petitioner must show (1) a case of statutory interpretation, (2) that is retroactive and could not have been invoked in the initial § 2255 motion, and (3) that the misapplied sentence presents an error sufficiently grave to be deemed a miscarriage of justice or a fundamental defect."); Brown v. Caraway , 719 F.3d 583, 586 (7th Cir. 2013) ("First, the prisoner must show that he relies on a statutory-interpretation case, rather than a constitutional case. Second, the prisoner must show that he relies on a retroactive decision that he could not have invoked in his first § 2255 motion. The third condition is that the sentence enhancement have been a grave enough error to be deemed a miscarriage of justice corrigible therefore in a habeas corpus proceeding." (quotations and citations omitted and alterations adopted)). See also United States v. Surratt , 797 F.3d 240, 274 (4th Cir. 2015) (Gregory, J., dissenting), reh'g en banc granted (Dec. 2, 2015) (" § 2255 is 'inadequate or ineffective' when the retroactively-applicable change in the law that the prisoner seeks to take advantage of occurs subsequent to his first § 2255 motion....[,] the asserted error represents a fundamental defect, [and] the prisoner cannot satisfy the gatekeeping provisions of § 2255 because he relies on a new rule that is not one of constitutional law." (quotation omitted and alterations adopted)).

The Legislative History of Section 2255 and the Supreme Court's Jurisprudence

The legislative history of § 2255 and the Supreme Court's habeas corpus jurisprudence confirm this view. Before § 2255 became law in 1948, federal prisoners who wanted to collaterally attack their conviction or sentence had to file a petition for habeas corpus in the district where they were in prison. This caused the few district courts located near federal prisons to be overwhelmed with habeas petitions. See United States v. Hayman , 342 U.S. 205, 213–15, 72 S.Ct. 263, 269–70, 96 L.Ed. 232 (1952). Congress enacted § 2255 to address this problem. The new statute "replaced traditional habeas corpus for federal prisoners ... with a process that allowed the prisoner to file a motion with the sentencing court." Boumediene v. Bush , 553 U.S. 723, 774, 128 S.Ct. 2229, 2264, 171 L.Ed.2d 41 (2008).

The Supreme Court has told us more than once that § 2255 was "designed to strengthen, rather than dilute, the writ's protections." Id. at 776, 128 S.Ct. at 2265. See also Davis v. United States , 417 U.S. 333, 343, 94 S.Ct. 2298, 2304, 41 L.Ed.2d 109 (1974) ("Th[e] [legislative] history makes clear that § 2255 was intended to afford federal prisoners a remedy identical in scope to federal habeas corpus."); id. at 344, 94 S.Ct. at 2304 ("Nowhere in the history of Section 2255 do we find any purpose to impinge upon prisoners' rights of collateral attack upon their convictions."); Hill v. United States , 368 U.S. 424, 427, 82 S.Ct. 468, 471, 7 L.Ed.2d 417 (1962) ( "[I]t conclusively appears from the historic context in which § 2255 was enacted that the legislation was intended simply to provide in the sentencing court a remedy exactly commensurate with that which had previously been available by habeas corpus in the court of the district where the prisoner was confined."); Hayman , 342 U.S. at 219, 72 S.Ct. at 272 ("[T]he sole purpose [of § 2255 ] was to minimize the difficulties encountered in habeas corpus hearings by affording the same rights in another and more convenient forum."). And beyond this admonition that § 2255 was not intended to weaken the Great Writ, it is critical that when Congress passed § 2255, it did not do away with traditional habeas corpus relief. Instead, it inserted the savings clause to preserve the habeas remedy for those instances in which § 2255"is inadequate or ineffective to test the legality of [a prisoner's] detention." 28 U.S.C. § 2255(e).

In 1996, Congress passed AEDPA to amend § 2255 by adding (among other things) the § 2255(h) limitations on filing more than one motion under that statute. At the same time, Congress did nothing to disturb the savings clause, and it remains a part of the law. It seems obvious that if Congress meant to bar all successive collateral attacks on convictions and sentences except for the two categories allowed by § 2255(h), it would have simply repealed the savings clause. It did not. I say AEDPA's narrowing of the availability of the § 2255 remedy only heightens the importance of the savings clause, whose express purpose is to ensure that, in every case, federal collateral review remains "[ ]adequate [and] [ ]effective." The Supreme Court told us in Boumediene —which was decided after AEDPA—that the purpose of the savings clause is to "provid[e] that a writ of habeas corpus would be available if the alternative process proved inadequate or ineffective." Boumediene , 553 U.S. at 776, 128 S.Ct. at 2265. So when there are "challenges to both convictions and sentences that as a structural matter cannot be entertained by use of the 2255 motion," § 2255 is "inadequate or ineffective." Webster v. Daniels , 784 F.3d 1123, 1139 (7th Cir. 2015).

The Importance of the Savings Clause Where Section 2255 Is "Inadequate or Ineffective"

There is clearly a gap in the protections offered by § 2255 in this circumstance: when the Supreme Court interprets a statute in a way that shows a prisoner's conviction or sentence was wrongly imposed, and that Supreme Court decision comes after the prisoner has already used up his first § 2255 proceeding. For Mr. McCarthan and other prisoners in his situation, the Supreme Court has given an interpretation of a statute that reveals they were sentenced to a term in prison longer than that authorized by Congress. And yet because the Supreme Court ruling comes after his first § 2255 proceeding is complete and because the decision is one of statutory (not constitutional) interpretation, he does not qualify to bring a second or successive motion under § 2255(h). So while Congress never gave the executive or judicial branches of government the power to keep Mr. McCarthan in jail this long, he has no remedy under § 2255. He has only the savings clause door to habeas corpus relief.

There are also other types of cases in which prisoners are serving sentences much longer than called for by law, with no remedy under § 2255, who should therefore be eligible for relief under the savings clause. For example, a prisoner who was convicted for conduct that the law does not in fact criminalize.⁷ Another example is prisoners who have been sentenced based on a mistaken application of the U.S. Sentencing Guidelines at the time when sentencing judges were mandated by law to follow them. See Brown v. Caraway , 719 F.3d 583, 587–88 (7th Cir. 2013) (holding that the savings clause allows a prisoner to challenge his detention when a retroactive statutory-interpretation decision reveals the prisoner was sentenced based on an erroneous application of the mandatory Sentencing Guidelines, even where the sentence does not exceed the statutory maximum); Gilbert , 640 F.3d at 1330–36 (Martin, J., dissenting).

7 Every circuit to have considered the issue—except for the Tenth Circuit and now this Circuit—has concluded that, at the least, the savings clause allows a prisoner to challenge his detention when a retroactive statutory-interpretation decision from the Supreme Court shows that the prisoner was convicted for conduct that the law does not in fact make criminal. See Trenkler v. United States , 536 F.3d 85, 99 (1st Cir. 2008) ; Poindexter v. Nash , 333 F.3d 372, 378 (2d Cir. 2003) ; In re Dorsainvil , 119 F.3d 245, 251–52 (3d Cir. 1997) ; In re Jones , 226 F.3d 328, 333–34 (4th Cir. 2000) ; Reyes–Requena v. United States , 243 F.3d 893, 903–04 (5th Cir. 2001) ; Wooten v. Cauley , 677 F.3d 303, 307–08 (6th Cir. 2012) ; Brown v. Caraway , 719 F.3d 583, 586–87 (7th Cir. 2013) ; Marrero v. Ives , 682 F.3d 1190, 1192, 1194–95 (9th Cir. 2012) ; In re Smith , 285 F.3d 6, 8 (D.C. Cir. 2002).

The Supreme Court has been clear that decisions "narrow[ing] the scope of a criminal statute by interpreting its terms" are given retroactive effect "because [such decisions] necessarily carry a significant risk that a defendant stands convicted of an act that the law does not make criminal or faces a punishment that the law cannot impose upon him." Schriro v. Summerlin , 542 U.S. 348, 351–52, 124 S.Ct. 2519, 2522–23, 159 L.Ed.2d 442 (2004) (quotation omitted). When the Supreme Court interprets a statute and applies its ruling retroactively, but a prisoner is barred from relying on that interpretation merely because the Supreme Court decided the case after his first § 2255 proceeding was done, § 2255 has certainly "proved inadequate or ineffective," Boumediene , 553 U.S. at 776, 128 S.Ct. at 2265, within the meaning of the savings clause. See Unthank v. Jett , 549 F.3d 534, 536 (7th Cir. 2008) (noting that the savings clause is available where "a glitch in § 2255 prevents application to [a petitioner's] situation of a retroactive decision of the Supreme Court").

The Majority says my reading of the savings clause would allow prisoners to make an end-run around the limitations on successive motions in § 2255(h). Maj. Op. at 1089–92. I say my reading just gives effect to the words Congress wrote. It is critical to remember that Congress preserved the savings clause as an avenue of relief for prisoners even as it passed strict restrictionson the filing of successive § 2255 motions. Under the rule the Majority adopts today, so long as the prisoner had a formal chance to raise his claim in a § 2255 motion—whether the court's ruling on that claim was right or wrong—the § 2255 proceeding is deemed "[ ]adequate [and] [ ]effective." That means the prisoner can never file another collateral attack on his sentence unless he can meet one of § 2255(h)'s two exceptions to the successive-motions bar. This, of course, reads the savings clause right out of the statute. As I have said before, "[b]y grafting the requirements of § 2255(h) onto the savings clause, the Majority has stripped that clause of any independent meaning." Gilbert , 640 F.3d at 1333 (Martin, J., dissenting).

As lawyers, we're taught that an interpretation rendering a statutory clause meaningless violates the "cardinal principle of statutory construction": that we must "give effect, if possible, to every clause and word of a statute." Williams v. Taylor , 529 U.S. 362, 404, 120 S.Ct. 1495, 1519, 146 L.Ed.2d 389 (2000). And this bromide takes on real significance when we use it to interpret a statute that governs habeas jurisdiction, because it affects so many real people who may be wrongly imprisoned. The Supreme Court has admonished us there is a "longstanding rule requiring a clear statement of congressional intent to repeal habeas jurisdiction." I.N.S. v. St. Cyr , 533 U.S. 289, 298, 121 S.Ct. 2271, 2278, 150 L.Ed.2d 347 (2001). No one writing on the other side of this issue has pointed to any indication—much less a clear statement—from Congress that it intended for § 2255(h) to repeal the savings clause of § 2255(e). So the Majority's reading should not stand. See Boumediene , 553 U.S. at 738, 128 S.Ct. at 2243 ("Congress should 'not be presumed to have effected such denial of habeas relief absent an unmistakably clear statement to the contrary.' " (quoting Hamdan v. Rumsfeld , 548 U.S. 557, 575, 126 S.Ct. 2749, 2764, 165 L.Ed.2d 723 (2006) (alteration adopted))).

Federal judges wield enormous power. But we are human beings who make mistakes. The Majority's interpretation of § 2255 leaves federal judges unaccountable when we wield our power to take away people's liberty for longer than the law allows. This is particularly striking here, where both of the other branches of government make our mistake clear. The Legislative branch passed a law allowing Mr. McCarthan's crime to be punished for up to ten years in prison, but no more. The Executive branch rejected the position the Majority takes here—to the extent that our Court had to bring in another lawyer to even advocate for the position it adopts today. And now the Majority, on behalf of the Judicial branch, has made a rule that prevents federal judges from correcting an illegal sentence. A system of government set up with branches to check and balance each other simply should not work this way. No one branch should be able to insulate its mistakes from its own review, much less the review of the other branches. Most anyone performs better, day in and day out, when they know they can be called to account for getting it wrong. Federal judges are no different.

II. THE MERITS OF MR. MCCARTHAN'S PETITION

In the midst of all of this debate about these statutes, it is important to now return to Mr. McCarthan's case. I agree with Judge Rosenbaum that Mr. McCarthan's case should be remanded to the District Court. However, I write separately because I believe the law limits what the District Court can do on remand. My understanding of the law tells me Mr. McCarthan is eligible for habeas relief under the savings clause, so I turn to the merits of his habeas petition. Mr. McCarthan argues he was sentenced to a term of imprisonment beyond that authorized by law because he does not have the three predicate felony convictions necessary to support the ACCA enhancement. He is right, and I would grant him relief.

Throughout its prosecution of Mr. McCarthan for being a felon in possession of a firearm, the government pointed to three (and only three) prior convictions as predicates for the ACCA enhancement. Those were: (1) a 1987 conviction in Florida for possession of cocaine with intent to sell or deliver; (2) a 1992 conviction in Florida for escape; and (3) a 1994 conviction in Florida for third-degree murder. Only these prior convictions were listed in the indictment. And these convictions were the only ones offered at Mr. McCarthan's guilty plea hearing to justify a 15-year minimum sentence under ACCA. Again when Mr. McCarthan was sentenced, the government mentioned no other convictions as qualifying him for an ACCA sentence.

Everyone agrees that Mr. McCarthan's escape conviction would not be accepted as a valid ACCA predicate for him if he were sentenced today. Both the Supreme Court and this Circuit have said so. Lee , 586 F.3d at 874 ; Chambers , 555 U.S. at 122, 129 S.Ct. at 687. This means, even if we assume that Mr. McCarthan's two remaining prior convictions properly support his longer sentence, all we have is two ACCA predicates. This is one short of the number of convictions required to keep Mr. McCarthan behind bars for more than 10 years. Because escape does not qualify as one of the three convictions required for an ACCA sentence of longer than ten years, and because the government offered no proof of any prior convictions other than the three it listed in Mr. McCarthan's indictment, Mr. McCarthan is being held in violation of § 924(a)(2), which sets a ten year limit on his prison term.

When the government asks a court to give a person a sentence above the term the statute sets as a limit, the government bears the burden of proving the longer sentence is proper under the law and the facts of the case. Lee , 586 F.3d at 866 ("The [government] bears the burden of proving that a sentencing enhancement under the ACCA is warranted."); see also United States v. Young , 527 F.3d 1274, 1277 (11th Cir. 2008). Here, the government asked for an enhanced sentence for Mr. McCarthan based solely on the three convictions listed in Mr. McCarthan's indictment. The government never mentioned any other basis for an enhanced sentence.

It is true that the PSR listed two other earlier felony convictions for Mr. McCarthan, those being two 1988 Georgia convictions for possession of cocaine. It is also true that, at sentencing, Mr. McCarthan did not object to any of the prior convictions in the PSR, and that the District Court adopted the facts stated in the PSR. But the PSR did not identify which convictions qualified Mr. McCarthan for an ACCA enhancement. So when he did not contest the PSR, Mr. McCarthan conceded only that these Georgia cocaine convictions existed . Whether these convictions qualify as "serious drug offenses" under ACCA is a separate question, and one that was never even discussed at any court proceeding that resulted in his 211-month sentence. The government never mentioned them. The sentencing judge never mentioned them. I write separately to reject any idea that it was incumbent upon Mr. McCarthan to interrupt his sentencing hearing, a time where he was no doubt nervously awaiting to hear his fate, to bring up these other convictions that no one else thought worthy of mention. He simply had no burden to disprove something the government never sought to prove in the first place. To place that burden on him would surely turn the sentencing process on its head.

To place that burden on him also defies our own precedent. Our Court has a waiver rule that says where the government never told the District Court (or for that matter the defendant being sentenced) that a particular conviction is a reason to impose a longer sentence, the government is barred on appeal from arguing that the previously unmentioned felony can now take the place of a conviction that was relied upon by the sentencing court, but which no longer supports the sentence. In Bryant , this Court considered and rejected the government's effort to bring up on appeal new bases for Mr. Bryant's longer sentence, when the reasons it gave at the time of his sentencing no longer supported the sentence he got. In Bryant , like here, the government sought to substitute a prior burglary conviction for a concealed-firearm conviction that no longer worked, when "[a]t no time during Bryant's direct criminal proceedings did the government ever rely on the burglary conviction as a predicate felony for § 924(e) purposes." 738 F.3d at 1279. We "den[ied] the government's request to substitute the burglary conviction" because "the government waived this burglary issue at the initial sentencing." Id. See also United States v. Canty , 570 F.3d 1251, 1257 (11th Cir. 2009) (stating that, while the government is "entitled to an opportunity to offer evidence and seek rulings from the sentencing court in support of an enhanced sentence," the government is "entitled to only one such opportunity"). Like Bryant , the government in this case "never suggested at any point" prior to collateral review that Mr. McCarthan's Georgia cocaine convictions could serve as ACCA predicates. 738 F.3d at 1279. The "government cannot offer for the first time on appeal a new predicate conviction in support of an enhanced ACCA sentence." United States v. Petite , 703 F.3d 1290, 1292 n.2 (11th Cir. 2013). This is as it should be, because our Court has never allowed criminal defendants to contest their harsh sentence on appeal for reasons they had not presented to the sentencing court. I cannot sanction applying different rules to opposing parties appearing in this Court in one and the same proceeding.

I would grant Mr. McCarthan relief and send his case to the District Court with direction that he be resentenced to a term of no more than ten years. Mr. McCarthan has, of course, already served more than ten years in the penitentiary. I respectfully dissent from this Court's treatment of Mr. McCarthan, as well as its remaking of our law as it governs habeas corpus for those sentenced in this Circuit.

ROSENBAUM, Circuit Judge, dissenting:

I agree with the Majority that we incorrectly interpreted 28 U.S.C. § 2255(e) on at least five occasions: Wofford v. Scott , 177 F.3d 1236 (11th Cir. 1999), Gilbert v. United States , 640 F.3d 1293 (11th Cir. 2011) (en banc), Williams v. Warden, Federal Bureau of Prisons , 713 F.3d 1332 (11th Cir. 2013), Bryant v. Warden, FCC Coleman-Medium , 738 F.3d 1253 (11th Cir. 2013), and Mackey v. Warden, FCC Coleman-Medium , 739 F.3d 657 (11th Cir. 2014). And today, unfortunately, makes a sixth.

Though the Majority is right when it concludes that the existence of adverse circuit precedent on a prisoner's claim has no relevancy to whether a second or successive claim may be brought under § 2255(e), the Majority's analysis is not itself faithful to the text of § 2255(e)'s so-called saving clause,¹ does not recognize the crucial constitutional-failsafe purpose that the saving clause serves, and does not acknowledge the role that the Suspension Clause plays in determining whether a second or successive claim may proceed under the saving clause. As a result, the Majority misses the fact that § 2255(e) must allow for consideration of second or successive claims that rely on a retroactively applicable new rule of statutory law.

The saving clause serves as a failsafe mechanism to protect § 2255 from unconstitutionality by providing a substitute remedy for habeas corpus relief that § 2255 otherwise precludes but the Suspension Clause may require.² And since the Suspension Clause exists to protect habeas corpus, the Suspension Clause demands, at a minimum, the availability of habeas corpus relief to redress federal detention when it violates the very doctrinal underpinnings of habeas review.

Habeas review, in turn, finds its doctrinal underpinnings in the doctrine of separation of powers and the principle of limited government powers. See Bousley v. United States , 523 U.S. 614, 620-21, 118 S.Ct. 1604, 140 L.Ed.2d 828 (1998) (characterizing separation-of-powers concerns as "the doctrinal underpinnings of habeas review"); see also Welch v. United States , –––U.S. ––––, 136 S.Ct. 1257, 1268, 194 L.Ed.2d 387 (2016) (equating the principle of limited government powers with separation-of-powers concerns in federal habeas jurisprudence). So detention that violates the separation-of-powers doctrine or the principle of limited government powers necessarily tramples upon the doctrinal underpinnings of habeas review. See Bousley , 523 U.S. 614, 118 S.Ct. 1604 ; Welch , 136 S.Ct. 1257.

To remedy this affront to habeas corpus, new rules of statutory and constitutional interpretation that reveal detention in violation of the separation of powers or the principle of limited government powers are retroactively applicable on federal collateral review. In other words, these new rules of constitutional and statutory law are retroactively applicable on federal collateral review because the doctrinal underpinnings of habeas corpus—and therefore the Suspension Clause—require that they be.

And the very same concepts that, under the Suspension Clause, demand the retroactivity of new rules of constitutional or statutory law on initial collateral review—the separation-of-powers doctrine and the principle of limited government powers—apply with equal force in the context of second or successive claims for collateral review based on a previously unavailable retroactively applicable rule of constitutional or statutory law. When a prisoner is detained in violation of the separation of powers or the principle of limited government, the violation does not somehow become less significant simply because the Supreme Court does not recognize the violation by issuance of a new retroactively applicable rule of law until after the prisoner's initial § 2255 claim has been resolved.

Indeed, § 2255(h)(2) implicitly recognizes this fact as it pertains to second or successive claims based on a new retroactively applicable rule of constitutional law. But since § 2255 does not authorize second or successive claims based on a retroactively applicable new rule of statutory law though the Suspension Clause requires courts' consideration of such claims when a prisoner has not previously had a meaningful opportunity to have had such claims heard, the saving clause necessarily must allow these claims in order to save § 2255 from unconstitutionality.

Dan McCarthan's claim relies on a new retroactively applicable rule of statutory law. So I would reverse the district court's dismissal of his petition and remand for consideration of the merits.

This first section of this dissent explains why the text of § 2255 and Supreme Court jurisprudence on habeas corpus, the Suspension Clause, and retroactivity necessarily require that the saving clause allow for consideration of second or successive claims based on a new retroactively applicable rule of statutory law. Part II of the dissent addresses the Majority's criticism of the theory I espouse in Part I. In Part III, I explore why the Majority's construction of the saving clause cannot be correct under the statutory text and Supreme Court precedent. And in Part IV, I respond to Chief Judge Carnes's concurrence.

I.

To assess whether the saving clause requires consideration of McCarthan's second or successive claim, we must answer two questions: first, does the saving clause permit at least some second or successive claims? And second, if so, does the saving clause allow second or successive claims that, like McCarthan's, are based on a new retroactive rule of statutory construction that, if applicable, would mean that the applicant has been imprisoned beyond valid congressional authorization? By itself, the statutory language of the saving clause tells us the answer to the first question is "yes." To resolve the second, we must consult the statutory language and the Supreme Court's jurisprudence on habeas corpus. Together, they reveal that the answer to the second question is also "yes."

A. The language of the saving clause necessarily contemplates that the saving clause will be used to bring at least some types of second or successive claims.

In all cases of statutory construction, we start our analysis by examining the language of the statute for a "plain and unambiguous meaning with regard to the particular dispute in the case." Barnhart v. Sigmon Coal Co., Inc. , 534 U.S. 438, 450, 122 S.Ct. 941, 151 L.Ed.2d 908 (2002) (citation and internal quotation marks omitted). Our inquiry ends here as well if the statutory language is "unambiguous and the statutory scheme is coherent and consistent." Id. (citation and internal quotation marks omitted). In this case, we can answer our first question—whether the saving clause allows for consideration of second or successive claims of at least some type—solely by consulting the statutory language.

The saving clause, in the context of § 2255(e), provides,

An application for a writ of habeas corpus in behalf of a prisoner who is authorized to apply for relief by motion pursuant to this section, shall not be entertained if it appears that the applicant has failed to apply for relief, by

motion, to the court which sentenced him, or that such court has denied him relief , unless it also appears that the remedy by motion is inadequate or ineffective to test the legality of his detention.

28 U.S.C. § 2255(e) (emphasis added). The words "such court" refer to "the court which sentenced him," so the words "such court has denied him relief" unambiguously contemplate that a prisoner previously made at least a first § 2255 motion,³ and his sentencing court denied it. That means that the claim that any such prisoner seeks to bring under the saving clause necessarily must be a second or successive claim. By its language, then, the saving clause specifically requires courts to consider a prisoner's second or successive claim when "it also appears that the remedy by [§ 2255 ] motion is inadequate or ineffective to test the legality of [the prisoner's] detention." Id.

Any reading of the saving clause that completely precludes courts from considering second or successive claims can achieve that result only by ignoring the language "such court has denied him relief" and its natural meaning. But the court has a "duty to give effect, if possible, to every clause and word of a statute." Duncan v. Walker , 533 U.S. 167, 174, 121 S.Ct. 2120, 150 L.Ed.2d 251 (2001) (citation and internal quotation marks omitted). When we do that in this case, we cannot escape the conclusion that the language of the saving clause plainly envisions consideration of at least some second or successive claims.⁴

B. Under the saving clause, a petitioner may bring a second or successive claim based on a new retroactively applicable rule of statutory law that means that his sentence exceeds what Congress has validly authorized.

Since the saving clause allows at least some second or successive claims, the question is, "Which ones?" When we view the terms of the saving clause in the light of Supreme Court precedent, the answer becomes clear. The saving clause allows for two categories of claims: (1) those that, though permissible under § 2255's provisions other than the saving clause, cannot, for practical and logistical reasons, be brought under those provisions, and (2) those that cannot otherwise be brought under the other parts of § 2255 and that are constitutionally required to be considered, including those that assert a prisoner is detained in violation of the government's, or a branch of the government's, powers, as supported by a retroactively applicable new rule of substantive law.

To identify the particular second or successive claims for which the saving clause requires consideration, we must focus on, in particular, three parts of the language of § 2255(e) : (1) "a prisoner who is authorized to apply for relief by motion pursuant to this section"; (2) "legality of his detention"; and (3) "inadequate or ineffective to test." Like a series of filters, each phrase limits the preceding universe of claims, yielding a successively smaller universe. So for a claim to be entitled to consideration under the saving clause, it must pass through all three filters.

1. "a prisoner who is authorized to apply for relief by motion pursuant to this section"

By making the saving clause applicable to "a prisoner who is authorized to apply for relief by motion pursuant to this section," the first filter the saving clause imposes restricts its availability to sentenced federal prisoners who raise a type of claim that § 2255(a) permits. The unambiguous statutory language of § 2255 dictates this construction.

To explain why, we begin by evaluating § 2255 for any language that authorizes a petitioner to apply for relief. A review of § 2255's language reveals that the only parts of it that "authorize" a "prisoner" to do anything include subsections 2255(a) and (d). Subsection (a) provides,

A prisoner in custody under sentence of a court established by Act of Congress claiming the right to be released upon the ground that the sentence was imposed in violation of the Constitution or laws of the United States, or that the court was without jurisdiction to impose such sentence, or that the sentence was in excess of the maximum authorized by law, or is otherwise subject to collateral attack, may move the court which imposed the sentence to vacate, set aside or correct the sentence.

28 U.S.C. § 2255(a) (emphasis added). By directing that a "prisoner ... may move ," subsection (a) plainly authorizes a prisoner to apply for relief under the circumstances set forth in subsection (a). Id. (emphasis added).

Similarly, subsection (d) states that "[a]n appeal may be taken to the court of appeals from the order entered on the motion as from a final judgment on application for a writ of habeas corpus." Id. § 2255(d) (emphasis added). Under our system, a losing litigant may take an appeal where permitted. So like subsection (a), subsection (d) authorizes an applicant to take action.

But none of the remaining parts of § 2255"authorize" a "prisoner" to do anything. Rather, they give processing instructions to the court. Subsection (b), for example, states, in relevant part,

Unless the motion and the files and records of the case conclusively show that the prisoner is entitled to no relief, the court shall ... [engage in various actions]. If the court [makes certain findings], the court shall vacate and set the judgment aside and shall [take appropriate corrective action].

Id. § 2255(b) (emphasis added). Likewise, subsection (c) provides that "[a] court may entertain and determine such motion without requiring the production of the prisoner at the hearing." Id. § 2255(c) (emphasis added). And subsection (g) enables the court to appoint counsel for proceedings under § 2255. See id. § 2255(g) ("[I]n all proceedings brought under this section, ... the court may appoint counsel ....") (emphasis added). Subsection 2255(h) tells the court how to process a second or successive motion. See id. § 2255(h) ("A second or successive motion must be certified ... by a panel of the appropriate court of appeals ....") (emphasis added). All of these parts of § 2255 instruct a court on how to handle a § 2255 application.Finally, subsection (f), AEDPA's statute of limitations, is either a jurisdictional requirement for the reviewing court or an affirmative defense for the defendant, depending on which circuit construes the provision. Compare , e.g. , Williams , 713 F.3d at 1338-40,⁵ with Acosta v. Artuz , 221 F.3d 117, 121-22 (2d Cir. 2000). Either way, it does not purport to authorize a prisoner to do anything. A jurisdictional provision empowers a court to hear a case, while an affirmative defense is a tool that a defendant may use to bar suit.

In short, only subsections (a) and (d) authorize a prisoner to take action.⁶ So we must review those subsections to determine which prisoners subsections (a) and (d) permit to apply for relief under § 2255.

6 The Majority argues that subsections (f) and (h) also "authorize" a prisoner to apply for relief. Based on the language of these sections, I respectfully disagree. But even if the Majority is right about that, it would have no impact on the ultimate conclusion that the saving clause requires consideration of second or successive claims that are based on a retroactively applicable new rule of statutory law. See infra at 1150–51 & 1151 n.22. Ironically, though, if the Majority is correct and subsections (f) and (h) also "authorize" a prisoner to apply for relief under § 2255, that fact would undermine some of the Majority's criticism of my theory in ways additional to those that exist if subsections (f) and (h) do not "authorize" a prisoner to apply for relief under § 2255. See id.; see also id. at 1146.

Beginning with subsection (d), as it pertains to prisoners, that subsection authorizes only appeals from denied claims brought under subsection (a). As a result, it does not expand the category of prisoners "authorized to apply for relief by motion pursuant to [§ 2255 ]" beyond what subsection (a) provides.

I therefore turn to subsection (a). By its terms, subsection (a) allows an applicant meeting four qualifications to seek relief under § 2255 ("A prisoner ... may move....").

First, the language requires an applicant to be "[a] prisoner in custody." 28 U.S.C. § 2255(a). That requirement is self-explanatory.

Second, the language "under sentence of a court established by Act of Congress" means that the prisoner must be a federal prisoner. See id.

Third, not just any federal prisoner may apply for relief under § 2255. Only federal prisoners who have been "sentence[d]" by a federal court are eligible to seek relief under the statute. See id. ("[a] prisoner ... under sentence ...") (emphasis added). So, for example, a pretrial detainee may not use § 2255 to seek relief.

And last, under subsection (a)—and therefore under subsection (e)—only those sentenced federal prisoners "claiming the right to be released upon [a] ground [that subsection (a) specifies]"—"the right to be released upon the ground that the sentence was imposed in violation of the Constitution or laws of the United States, or that the court was without jurisdiction to impose such sentence, or that the sentence was in excess of the maximum authorized by law, or is otherwise subject to collateral attack," id. —are "authorized" to bring a § 2255 motion.

The clause "claiming the right to be released upon the ground[s] [articulated in subsection (a) ]"⁷ is a restrictive clause that modifies the subject clause in subsection (a)—"[a] prisoner in custody under sentence of a court established by Act of Congress." Restrictive clauses limit and define the subjects they modify, and unlike non-restrictive clauses, they are not set off by commas. See Strunk & White, supra , at 16. Significantly, the "claiming" clause in subsection (a) is not separated from the subject clause by a comma and is plainly intended as a restrictive clause. It therefore limits and defines the "prisoner in custody" clause. As Justice Scalia and Bryan Garner have noted, Congress is "presumed to be grammatical in [its] compositions." Antonin Scalia & Bryan A. Garner, Reading Law: The Interpretation of Legal Texts 140 (2012). For this reason, only those sentenced federal prisoners in custody who are claiming one of the specific violations set forth in subsection (a) "may move" for relief—and are therefore "authorized to apply for relief"—by § 2255 motion.

7 The words "who is" are understood at the beginning of the clause "claiming the right to be released upon the ground[s] [articulated in subsection (a) ]." See, e.g., William Strunk Jr. & E.B. White, The Elements of Style 16 (4th ed. 2000) (providing as an example of a restrictive clause, "People sitting in the rear couldn't hear.").

The Majority incorrectly contends that Congress intended under subsection (e) to include as prisoners "authorized" to bring a § 2255 motion, those prisoners bringing any type of collateral claim at all—not just the challenges to sentences that subsection (a) allows. See Maj. Op. at 1092–93. That cannot be correct for three independent reasons. First, as explained above, the text and grammatical structure of the saving clause do not bear the Majority's proposed interpretation. Second, a comparison of the wording of subsections (a) and (e) does not support the Majority's theory. And finally, the function of § 2255 has only ever dealt with federal prisoners' sentencing claims and not indiscriminately with all kinds of collateral claims.

Turning to the second reason, if Congress had intended under subsection (e), as the Majority suggests, to include as prisoners "authorized" to bring a § 2255 motion, those prisoners bringing any kind of collateral claim, Congress had a ready way of expressing that—which it chose not to use. In subsection (e), Congress could have relied on the phrase "[a] prisoner in custody under [order] of a court established by Act of Congress," similar to what it employed in subsection (a) before limiting that phrase with a laundry list of specific permissible claims. See 28 U.S.C. § 2255(a), (e). Had Congress done so, it would have authorized consideration of any collateral claim of a federal prisoner—not just collateral claims relating to sentencing.

But Congress did not do that.

Instead, it relied on a different and slightly longer phrasing. Under subsection (e) as Congress actually enacted it, that provision allows for consideration of an application from only "a prisoner who is authorized to apply for relief by motion pursuant to this section." Id. § 2255(e). So "a prisoner who is authorized to apply for relief by motion pursuant to [§ 2255 ]," id. must mean something different than "[a] prisoner in custody under [order] of a court established by Act of Congress," id. § 2255(a), because "[w]e generally seek to respect Congress' decision to use different terms to describe different categories of people or things." Mohamad v. Palestinian Auth. , 566 U.S. 449, 132 S.Ct. 1702, 1708, 182 L.Ed.2d 720 (2012).

And, indeed, the phrase does mean something different. "[A] prisoner who is authorized to apply for relief by motion pursuant to [§ 2255 ]," 28 U.S.C. § 2255(e), necessarily means a sentenced federal prisoner in custody who seeks relief on one of the claims specified in subsection (a)—that is, a federal prisoner in custody after sentencing, who is "claiming the right to be released upon the ground that the sentence was imposed in violation of the Constitution or laws of the United States, or that the court was without jurisdiction to impose such sentence, or that the sentence was in excess of the maximum authorized by law, or is otherwise subject to collateral attack," id. § 2255(a).

Returning to the language of subsection (e), it provides that the habeas petition of a prisoner "authorized to apply for relief" under § 2255"shall not be entertained if it appears that the applicant has failed to apply for relief, by [§ 2255 ] motion ... unless it also appears that the remedy by [§ 2255 ] motion is inadequate or ineffective to test the legality of his detention." Id. § 2255(e). This language expressly contemplates that the saving clause allows courts to consider habeas petitions of only those prisoners bringing one of the four types of claims articulated in § 2255(a), and only if "it also appears that the remedy by motion is inadequate or ineffective to test the legality of [the prisoner's] detention."

Not surprisingly, this straightforward interpretation is also entirely consistent with § 2255's intended function—to provide a more practical substitute remedy for habeas corpus in cases of federal prisoners who challenge their sentences. Congress enacted § 2255 in 1948 to address the problems created by large-scale administration of habeas corpus. Among other practical problems, it was not economical to haul multiple witnesses across the country for a hearing on a collateral challenge to a sentence in a forum where the prisoner had not been sentenced. See United States v. Hayman , 342 U.S. 205, 212-14, 72 S.Ct. 263, 96 L.Ed. 232 (1952). Doing so imposed both the expense of transporting the witnesses from their home district to the district where the prisoner was housed and the judicial cost of requiring a second judge to familiarize herself with the prisoner's case and sentencing. So Congress passed § 2255 as a habeas substitute that did not in any way limit the substantive scope of habeas but merely shifted the forum for cases involving sentenced federal prisoners challenging their sentences, to the district of the sentencing court, often a district different from the district of confinement. See id. at 219, 72 S.Ct. 263 ; Swain v. Pressley , 430 U.S. 372, 377-78, 97 S.Ct. 1224, 51 L.Ed.2d 411 (1977).

The practical concerns that motivated the enactment of § 2255 pertain to claims involving sentencing and related conviction challenges, so by its terms, § 2255 provides a substitute remedy for habeas corpus for only those collateral claims that raise sentencing and related conviction challenges. See 28 U.S.C. § 2255(a) (federal prisoners may bring claims under § 2255"claiming the right to be released upon the ground that the sentence was imposed in violation of the Constitution or laws of the United States, or that the court was without jurisdiction to impose such sentence , or that the sentence was in excess of the maximum authorized by law, or [that the sentence ] is otherwise subject to collateral attack, may move the court which imposed the sentence to vacate, set aside or correct the sentence ") (emphasis added).

Indeed, the same concerns of financial expense and cost in judicial economy do not apply in the context of other types of collateral claims, such as execution-of-sentence claims. In execution-of-sentence claims, witnesses are generally located in the district where the § 2241 claim is filed, and familiarity with the prisoner's underlying case and sentencing is not required to the same extent as in sentencing claims, if it is required at all.

So collateral claims that do not raise challenges to a prisoner's sentence have been able to be brought in habeas corpus under § 2241 ⁸ since its enactment at the same time as § 2255,⁹ and those claims have never been affected in any way by § 2255. See 28 U.S.C. § 2241 ; see also Antonelli v. Warden, U.S.P. Atlanta , 542 F.3d 1348, 1352 (11th Cir. 2008) ("challenges to the execution of a sentence, rather than the validity of the sentence itself, are properly brought under § 2241"). True, collateral claims attacking the validity of a federal conviction and sentence normally may not be brought in a habeas petition under § 2241—but only because § 2255 expressly carves out those specific claims from § 2241's authorization of courts' consideration of habeas claims. See 28 U.S.C. § 2255(e). Since § 2255 does not cover non-sentencing claims in the first place, there is no need for—and, indeed, it would make no sense for—the saving clause to exempt from § 2255's coverage collateral claims that do not raise sentencing challenges.

8 Section 2241"descends directly from § 14 of the Judiciary Act of 1789[,] [which authorized federal courts to issue writs of habeas corpus,] and the 1867 Act [that amended the Judiciary Act of 1789]." INS v. St. Cyr, 533 U.S. 289, 305 & n.25, 121 S.Ct. 2271, 150 L.Ed.2d 347 (2001).

9 Both statutes were enacted in 1948. Section 2241(e)(2) refers expressly to habeas petitions "relating to any aspect of the detention, transfer, treatment, ... or conditions of confinement." (emphasis added). Though it provides for no jurisdiction for such claims when they are raised by enemy combatants, the unambiguous negative implication is that courts have jurisdiction to entertain non-enemy-combatant prisoners' habeas petitions raising execution-of-sentence claims. See Scalia & Garner, supra, at 107-11.

To summarize, (1) the plain meaning and grammatical structure of the text of subsections (e) and (a); (2) the deliberate difference between the phrasing of "a prisoner who is authorized to apply for relief by motion pursuant to [§ 2255 ]," 28 U.S.C. § 2255(e), and "[a] prisoner in custody under sentence of a court established by Act of Congress," id. § 2255(a) ; and (3) the fact that § 2255 has only ever dealt with federal prisoners' sentencing claims and not execution-of-sentence or pretrial-detention claims, all demand the following conclusion: the first limitation subsection (e) unambiguously imposes on the availability of habeas-corpus relief through the saving clause requires that a petitioner be a sentenced federal prisoner in custody who is making a claim expressly authorized by subsection (a).

2. "legality of his detention"

Next, subsection (e) filters the universe of claims that a qualifying prisoner may otherwise raise under subsection (a), allowing through only those claims that test the "legality of [the applying prisoner's] detention." See 28 U.S.C. § 2255(e). The plain language of this phrase limits eligible claims to only those where the prisoner's success on his claim would result in a reduced period of detention.

For example, a prisoner may be sentenced to two or more concurrent terms of imprisonment. If that prisoner does not challenge the conviction or sentence that resulted in the longest period of imprisonment, he does not challenge the "legality of his detention." That's because even if that prisoner succeeds on his claim, he will remain legally detained for the exact same period for which he was to be detained before he filed his claim, since his unchallenged sentence requiring that will remain in force. In that instance, the saving clause does not reach the prisoner's § 2255(a) claim. See, e.g. , Brown v. Warden, FCC Coleman-Low , 817 F.3d 1278, 1284 (11th Cir. 2016) ("When a prisoner has only one conviction and sentence, his detention is legal as long as his sentence is legal. However, if a prisoner is serving multiple sentences, his detention may be legal even if one of his sentences is not."). So not all § 2255(a) claims necessarily challenge the "legality of [a prisoner's] detention." And only those that do survive § 2255(e)'s second filter.

3. "inadequate or ineffective to test"

Finally, we come to § 2255(e)'s third filter: "inadequate or ineffective to test." 28 U.S.C. § 2255(e). Congress's use of the disjunctive in the phrase "inadequate or ineffective" has significance. When Congress employs the word "or," "the words it connects are to be given separate meanings." Loughrin v. United States , ––– U.S. ––––, 134 S.Ct. 2384, 2390, 189 L.Ed.2d 411 (2014) (citation and internal quotation marks omitted). So in the saving clause, the words "inadequate" and "ineffective" have different and distinct meanings. And because these words are joined by "or," a prisoner must demonstrate that his claim satisfies only one of these standards (though some claims will satisfy both): the remedy by § 2255 must be either "inadequate" or "ineffective" for a prisoner's claim to pass through § 2255(e)'s third filter.

The Majority resists this intuitive interpretation, turning the natural meaning of "or" on its head. In the Majority's view, the saving clause uses "or" to "introduce[ ] a synonym or definitional equivalent." Maj. Op. at 1087–88 (citing Scalia & Garner, supra , at 122). So under the Majority's analysis, Congress intended for "inadequate" and "ineffective" to mean the same thing in the saving clause.

Notably, the "synonym-introducing or " exception that the Majority relies on "is typically set off by commas." Scalia & Garner, supra , at 122. But that is not the case with the words "inadequate or ineffective" in the saving clause.

So, to explain how it reaches this conclusion that the far less natural "synonym-introducing or " exception to the conjunctive/disjunctive canon applies to the "or" in the saving clause, the Majority simply states that "it is the better reading of the text when the terms share the same ordinary meaning." Maj. Op. at 1088. But this reasoning is circular: to decide the meanings of "inadequate" and "ineffective" in the saving clause, the Majority relies on the "synonym-introducing or " exception, which it, in turn, relies on because it finds that the words have the same meaning.

Habeas is a specialized area of the law, and in specialized areas of the law, words are often endowed with specialized meanings—creating "terms of art"—as in the case of the saving clause. Indeed, both "inadequate" and "ineffective" are used in habeas jurisprudence as legal terms of art. But the Majority does not even consider this fact.

And the Majority's application of the "synonym-introducing or " exception to the conjunctive/disjunctive canon cannot be correct for another reason: the Majority's interpretation of "inadequate" and "ineffective" as definitional equivalents does not account in any way for the saving clause's crucial constitutional-failsafe function. See infra at 1132–34.

Since the "or" in the saving clause indicates that "inadequate" and "ineffective" have different meanings from each other, we must consider what each word encompasses. We apply the meanings these words have acquired through Supreme Court habeas precedent. a. "inadequate ... to test"

Beginning with the term "inadequate," this term of art appears in the jurisprudence of equity, of which habeas jurisprudence is a part. See Boumediene , 553 U.S. at 780, 128 S.Ct. 2229 ("Habeas 'is, at its core, an equitable remedy' " (quoting Schlup v. Delo , 513 U.S. 298, 319, 115 S.Ct. 851, 130 L.Ed.2d 808 (1995) )). It is well established that a remedy at law is "inadequate" if it is not "as complete, practical and efficient as that which equity could afford." Terrace v. Thompson , 263 U.S. 197, 214, 44 S.Ct. 15, 68 L.Ed. 255 (1923).

Applying that concept in the context of the saving clause, and giving effect to a significant part of the congressional motivation behind § 2255's enactment (dealing with practical problems that arose under the pre-§ 2255 habeas regime, see supra at 1128), § 2255 is inadequate if practical considerations effectively or actually render the procedures § 2255 establishes unavailable for testing the legality of a prisoner's detention. So, for example, imagine a physically challenged or medically limited prisoner who, at the time that the saving clause was originally enacted in 1948, was housed in a different district from where he was sentenced and was a necessary witness in his collateral case, though he could not travel. One possible solution could involve allowing the prisoner to use the saving clause because, in that case, the review provisions of § 2255 might not be able to, as a practical matter, provide a remedy. The saving clause then opens the gateway to habeas corpus to allow such a prisoner access to collateral relief.¹⁰ When we apply these considerations in a case like McCarthan's, though, we see that § 2255 does not provide an "inadequate" remedy because the prisoner in such a case does not have the type of practical or logistical problems in using § 2255's remedy mechanism that the word "inadequate" contemplates.

10 The Majority relies on Prost v. Anderson, 636 F.3d 578, 588 (10th Cir. 2011), to argue that the remedy by § 2255 is inadequate where an otherwise-permissible § 2255 claim that challenges the legality of detention cannot proceed because the sentencing court has dissolved, as in the case of a military prisoner. Maj. Op. at 1092–93. Judge Jordan disagrees. Under the current state of the law, I do not believe that the answer to this question is clear. As far as I can tell, neither we nor the Supreme Court has yet determined whether military prisoners' habeas corpus petitions pass through the saving clause or whether instead they are authorized directly under § 2241. The answer to this question depends, in turn, on whether the military petitioner's claim is "authorized" by § 2255. See 28 U.S.C. § 2255(e). For if it is, the language of § 2255(e) requires him to bring his habeas petition pursuant to § 2255's strictures, meaning he must bring it under the saving clause. See id. But whether § 2255"authorize[s]" the military prisoner's claim—a question that turns on whether a military court is a "court established by Act of Congress" within the meaning of § 2255(a) —does not appear to be a settled matter. Judge Jordan refers to Prost 's citation of Ackerman v. Novak, 483 F.3d 647 (10th Cir. 2007), a case in which the Tenth Circuit held that, "because a military court martial is not a 'court of the United States' within the meaning of § 2244(a), the prisoner did not need to obtain circuit authorization to file a § 2241 habeas corpus petition." Jordan Op. at 1109–1110 (quoting Ackerman, 483 F.3d at 651-53 ). But that does not tell us whether a military tribunal is a "court established by Act of Congress" within the meaning of § 2255(a). And the Tenth Circuit in Ackerman found that "military justice courts are established by an Act of Congress," even though these courts are not considered "courts of the United States." 483 F.3d at 652. So Ackerman may provide some support for the proposition that military prisoners' habeas petitions must pass through the saving clause. Yet while the court in Ackerman found § 2255 unavailable, see id. at 649 n.2, the court did not pass on the applicability of the saving clause. In sum, I find the law unclear as to whether a prisoner in custody under sentence of a military court brings a § 2241 petition directly under § 2241 or whether that petition must first pass through § 2255(e). If the latter, then § 2255 is "inadequate" to test the legality of the petitioner's detention, given the practical difficulty created by the dissolution of the petitioner's sentencing court. It is likewise not clear to me, as it is not clear to Judge Jordan, whether the habeas petitions of prisoners convicted in territorial courts that no longer exist must pass through the saving clause. See Jordan Op. at 1110–1111. If the saving clause is necessary, however, then I suspect that it would be because § 2255 is "inadequate" for these petitioners as well, given the logistical conundrum posed by the non-existence of their sentencing courts. We need not resolve these vexing questions today. In any case, we know that in 1948, Congress was concerned with the practical and logistical problems attendant to the administration of habeas corpus, and Congress was likely mindful that such problems may continue to exist even after the enactment of § 2255. By including the saving clause in § 2255, Congress protected against that problem.

b. "ineffective to test"

On the other hand, § 2255's remedy, or procedures, are "ineffective to test" the legality of a prisoner's detention when the prisoner files a second or successive claim like McCarthan's, asserting that a retroactively applicable new rule of statutory law means that the prisoner's sentence exceeds what Congress has validly authorized.

The term "ineffective" is a term of art in Sixth Amendment claims, a frequent subject of habeas jurisprudence. In this context, "ineffective" means "constitutionally deficient," as in "ineffective assistance of counsel." See, e.g., Strickland v. Washington , 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) ; see also Powell v. Alabama , 287 U.S. 45, 71-72, 53 S.Ct. 55, 77 L.Ed. 158 (1932) (noting that "the right to have counsel appointed, when necessary, is a logical corollary from the constitutional right to be heard by counsel" and holding that "under the circumstances ..., the necessity of counsel was so vital and imperative that the failure of the trial court to make an effective appointment of counsel was likewise a denial of due process within the meaning of the Fourteenth Amendment.").

Though, like me, the Majority also looks to the context of "ineffective assistance of counsel" for instruction on the meaning of "ineffective," see Maj. Op. at 1088–89 (quoting Brown v. Caraway , 719 F.3d 583, 597 (7th Cir. 2013) (Easterbrook, C.J., concerning the circulation under Circuit Rule 40(e))), it nonetheless opines that this dissent "takes this analogy too far when it asserts that because the ineffective assistance of counsel creates a constitutional deficiency under the Sixth Amendment, the term 'ineffective' means 'constitutionally deficient' " in the saving clause. Id .

But the Majority offers no reason why that is so—especially considering that ineffective-assistance claims are nearly exclusively decided in the context of collateral review (particularly in the Eleventh Circuit). Instead, the Majority just conclusorily asserts that the meaning of "ineffective" in the saving clause cannot be "constitutionally deficient." I respectfully disagree.¹¹

11 As discussed below, if "ineffective" did not include the definition of "constitutionally deficient" in the context of the saving clause, the saving clause would not serve the failsafe-constitutional purpose that the Supreme Court has repeatedly relied upon to find § 2255 constitutional. See Boumediene, 553 U.S. at 776, 128 S.Ct. 2229 ("The [Supreme] Court placed explicit reliance upon [the saving clause] provisions in upholding [28 U.S.C. § 2255 and the District of Columbia equivalent of § 2255 ] against constitutional challenges." (citing Swain, 430 U.S. at 381, 97 S.Ct. 1224 ; Hayman, 342 U.S. at 223, 72 S.Ct. 263 )).

Section 2255 is "ineffective"—or constitutionally deficient—when it fails to allow for consideration of any claims authorized by § 2255(a) that the minimum constitutional requirements of habeas corpus that the Suspension Clause of the Constitution imposes, demand. As relevant here, the Suspension Clause requires that prisoners (1) have a "meaningful opportunity" (2) to have a court consider any claim that relies on a new retroactively applicable rule of law that reveals that a petitioner's sentence exceeds what Congress has validly authorized.

To explain why, we start by considering why the saving clause must authorize consideration of any claims that the Suspension Clause requires if such claims may not be reviewed under any other part of 28 U.S.C. § 2255.

The Suspension Clause of the Constitution provides, "The Privilege of the Writ of Habeas Corpus shall not be suspended, unless when in Cases of Rebellion or Invasion the public Safety may require it." U.S. Const. art. I, § 9, cl. 2. This provision constitutionally "secure[s] the writ [of habeas corpus] and ensure[s] its place in our legal system." Boumediene , 553 U.S. at 740, 128 S.Ct. 2229. As the Supreme Court has explained, "The Framers viewed freedom from unlawful restraint as a fundamental precept of liberty, and they understood the writ of habeas corpus as a vital instrument to secure that freedom." Id. at 739, 128 S.Ct. 2229.

Since the Suspension Clause protects the writ of habeas corpus under the Constitution, the Supreme Court has always construed the saving clause of § 2255 to ensure access to the writ of habeas corpus commensurate with what the Suspension Clause constitutionally may require, to the extent that the rest of § 2255 does not provide for such review.¹² See id. at 776, 128 S.Ct. 2229 ("The [Supreme] Court placed explicit reliance upon [the saving clause] provisions in upholding [28 U.S.C. § 2255 and the District of Columbia equivalent of § 2255 ] against constitutional challenges." (citing Swain , 430 U.S. at 381, 97 S.Ct. 1224 ; Hayman , 342 U.S. at 223, 72 S.Ct. 263 )). The Supreme Court has expressly warned that failure to interpret the saving clause in this way would raise "serious question[s] about the constitutionality of [§ 2255 ]." Id. (quoting Swain , 430 U.S. at 381, 97 S.Ct. 1224 (internal quotation marks omitted)). Put simply, the saving clause plays the critical role of a constitutional failsafe for § 2255.

12 When Congress initially passed § 2255, the statute had no numbered subsections. But the language of the saving clause—what is now located at § 2255(e) of the current version of the statute—appeared verbatim as part of the original enactment of § 2255. See Pub. L. No. 80-773, 62 Stat. 869, 968 (1948).

For this reason, the saving clause must require consideration of any second or successive claims that the Suspension Clause protects but the rest of § 2255 fails to permit. If the saving clause did not, § 2255 would violate the Suspension Clause.

So we must consider the constitutional demands of the Suspension Clause. As explained next, Supreme Court precedent shows that, at a minimum, the Suspension Clause requires that prisoners have a "meaningful opportunity" for judicial consideration of any claim that a court has imposed detention in violation of the government's powers, whether because of a violation of the doctrine of separation of powers or a violation of the principle that our government is one of limited powers. As a result, a prisoner may show that § 2255 is constitutionally deficient under the Suspension Clause if his claim meets two requirements: (1) the claim must assert that his detention violates the principles of separation of powers or limited government, or both; and (2) the prisoner must not have had a "meaningful opportunity" to have brought this claim previously.

i. At a minimum, the Suspension Clause requires consideration of claims alleging that a prisoner's imprisonment was imposed in excess of government powers.

We begin by considering the types of claims that the Suspension Clause requires be heard. As Supreme Court precedent demonstrates, the Suspension Clause demands consideration of claims raising challenges that a sentence was imposed in excess of a government branch's valid powers.

The Supreme Court has described the separation of powers as the "essential design of the Constitution."¹³ Boumediene , 553 U.S. at 745, 128 S.Ct. 2229. To help ensure the continuing vitality of our system of government, the Framers viewed the writ of habeas corpus, in turn, as "an essential mechanism in the separation-of-powers scheme." Id. at 743, 128 S.Ct. 2229 ; see also id. at 765, 128 S.Ct. 2229 ("the writ of habeas corpus is itself an indispensable mechanism for monitoring the separation of powers"). They similarly emphasized the important role that habeas corpus plays in "preserv [ing] limited government." Id. at 744, 128 S.Ct. 2229 (citing The Federalist No. 84 (Alexander Hamilton)).

13 And with good reason: "The accumulation of all powers, legislative, executive, and judiciary, in the same hands, whether of one, a few, or many, and whether hereditary, selfappointed, or elective, may justly be pronounced the very definition of tyranny." The Federalist No. 47 (James Madison), available at https://www.congress.gov/resources/display/content/The+Federalist+Papers (last accessed Jan. 3, 2017).

So, at its core, habeas corpus is about keeping government powers in constitutional check. It accomplishes this crucial function by requiring consideration of claims where a prisoner tests the legality of his imprisonment on the basis that, in jailing him, at least one of the branches of government violated the separation of powers or the principle of limited government by exceeding its constitutional powers.

The Suspension Clause, in turn, safeguards the writ of habeas corpus, so it necessarily constitutionally requires consideration of claims revealing that a branch of the government has exceeded its constitutional powers to the same extent that habeas does. Indeed, the Supreme Court has explained that the Suspension Clause "protects the rights of the detained by a means consistent with the essential design of the Constitution ... to maintain the 'delicate balance of governance' that is itself the surest safeguard of liberty." Boumediene , 553 U.S. at 745, 128 S.Ct. 2229 (quoting Hamdi v. Rumsfeld , 542 U.S. 507, 536, 124 S.Ct. 2633, 159 L.Ed.2d 578 (2004) (plurality opinion)). For this reason, the separation-of-powers doctrine and the principle of limited government powers "must inform the reach and purpose of the Suspension Clause." Id. at 746, 128 S.Ct. 2229 (emphasis added).

That brings us back to the critical role of the Suspension Clause in understanding the meaning of § 2255(e). As we have already established, § 2255(e) must allow for consideration of all § 2255(a) claims protected by the Suspension Clause but not otherwise permitted by § 2255. And the Suspension Clause demands, at a minimum, that claims that show that the government or a branch of it has acted in excess of its constitutional powers be cognizable under habeas. So the separation-of-powers doctrine and the closely related principle of limited government powers act as the twin Rosetta Stones we must use to reveal the claims under § 2255(e) for which § 2255 can be constitutionally deficient—or "ineffective"—under the Suspension Clause, to test the legality of detention.

When we view potential habeas-corpus claims through the prism of the doctrine of separation of powers and the principle of limited government powers, we can see that the Suspension Clause demands consideration of those claims that challenge the power of the government to impose detention, whether because the imprisoning branch has exceeded its constitutionally authorized powers or because the government as a whole lacks the constitutional powers to detain under the circumstances. So when § 2255(e) speaks of § 2255's "ineffective[ness] to test," it necessarily is concerned with the constitutional deficiency that can exist under the Suspension Clause when a prisoner seeks to test the authorization of the government, under the separation of powers and the principle of limited government powers, to impose detention, and § 2255 does not allow that claim to proceed.

In the context of the Judiciary, under the separation of powers and the principle of limited government powers, we may not sentence someone for a "crime" that no congressionally enacted statute actually criminalizes (or validly criminalizes), nor may we sentence a person to more time in prison than Congress has validly authorized. See United States v. Lanier , 520 U.S. 259, 267 n.6, 117 S.Ct. 1219, 137 L.Ed.2d 432 (1997) ("Federal crimes are defined by Congress, not the courts."); United States v. DiFrancesco , 449 U.S. 117, 139, 101 S.Ct. 426, 66 L.Ed.2d 328 (1980) ("[A] defendant may not receive a greater sentence than the legislature has authorized."). Congress—not the Judiciary—has the power to define crimes and their respective punishments. See United States v. Kebodeaux , ––– U.S. ––––, 133 S.Ct. 2496, 2503, 186 L.Ed.2d 540 (2013). So, as relevant in McCarthan's case, the "ineffective to test" filter permits consideration of those claims that challenge a sentence imposed beyond the Judiciary's constitutionally authorized powers. But, as we next discuss, it does so only when the prisoner has not previously had a "meaningful opportunity" to have his claim considered.

ii. The Suspension Clause entitles a prisoner to a "meaningful opportunity" to have his claim considered.

In addition to protecting, at a minimum, habeas claims challenging imprisonment in excess of the government's powers and the separation of powers, the Suspension Clause imposes another requirement: a prisoner must have a "meaningful opportunity" to present his claim. As the Supreme Court has explained, "We ... consider it uncontroversial ... that the privilege of habeas corpus entitles the prisoner to a meaningful opportunity to demonstrate that he is being held pursuant to 'the erroneous application or interpretation' of relevant law." Boumediene , 553 U.S. at 779, 128 S.Ct. 2229 (emphasis added) (quoting St. Cyr , 533 U.S. at 302, 121 S.Ct. 2271 ).

And because the Suspension Clause requires the availability of habeas corpus for cases of detention resulting from a branch's overreaching its powers, § 2255 is a constitutionally deficient substitute for habeas corpus—and is therefore "ineffective to test the legality of ... detention"—if it does not provide a "meaningful opportunity" for a prisoner to "test" the authorization of the government, under the separation of powers and the principle of limited government powers, to imprison him.

As Supreme Court precedent shows, a remedy does not provide a prisoner with a "meaningful opportunity" and is therefore constitutionally deficient if it does not allow a prisoner to present a collateral challenge at a meaningful time. That is precisely the case under § 2255 (without reference to the saving clause) when a prisoner tries to present a second or successive claim based on a retroactively applicable new rule of statutory law.

When we're talking about the Suspension Clause, timing matters. As the Supreme Court's collateral-review retroactivity jurisprudence emphatically demonstrates, the timing of a remedy's availability can be the difference between constitutional deficiency and constitutional sufficiency. Indeed, the very existence of retroactivity doctrine itself is a shrine to the concept that a prisoner must be able to present his Suspension-Clause-required claim at a meaningful time.

When a new rule is retroactively applicable, a prisoner may make a new claim based upon it even if he raised the same issue on direct appeal and lost and his conviction became final before the Supreme Court announced the new rule. If timing were irrelevant to habeas-corpus jurisprudence, the Supreme Court would have had no reason to develop retroactivity analysis. As long as nothing prevented a petitioner from having raised an issue at trial or in his direct appeal, that would have been enough to satisfy habeas corpus concerns that a litigant have had a meaningful opportunity to present his argument—even if, after the prisoner's conviction became final, the Supreme Court determined a new substantive rule that, had it been issued earlier, would have required granting the prisoner's challenge. On collateral review, then, we would hear only claims based on arguments that, by their nature, generally could not have been raised on direct review—arguments like ineffective assistance of counsel, see Kimmelman v. Morrison , 477 U.S. 365, 106 S.Ct. 2574, 91 L.Ed.2d 305 (1986), and discovery of new, previously unavailable evidence showing actual innocence.

But that's not how collateral review works.

Instead, the Supreme Court has spent considerable time and effort developing the retroactivity framework. See, e.g. , Mackey v. United States , 401 U.S. 667, 675-702, 91 S.Ct. 1160, 28 L.Ed.2d 404 (1971) (Harlan, J., concurring); Teague v. Lane , 489 U.S. 288, 299-316, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989) (plurality opinion); Bousley , 523 U.S. at 616-24, 118 S.Ct. 1604 ; Welch , 136 S.Ct. at 1260-68. It has done this because "the 'retroactivity' of a new constitutional rule [is] a function of the scope and purposes of the habeas corpus writ." Mackey , 401 U.S. at 684, 91 S.Ct. 1160 (Harlan, J., concurring). In other words, retroactivity exists because habeas corpus and thus the Suspension Clause constitutionally require it .

Under the retroactivity framework, a prisoner may obtain relief on collateral review of a new claim raising an issue that was denied on direct review, where the Supreme Court has announced a new rule of substantive law after the prisoner's conviction became final.¹⁴ Bousley , 523 U.S. at 620-21, 118 S.Ct. 1604 ; Welch , 136 S.Ct. at 1264-67.

14 Prisoners may also obtain relief on collateral review of claims based on a "new watershed rule[ ] of criminal procedure." Welch, 136 S.Ct. at 1264 (citation and internal quotation marks omitted). These are procedural rules that implicate "the fundamental fairness and accuracy of the criminal proceeding." Id. (citation and internal quotation marks omitted). To date, the Supreme Court has identified only the right to counsel as falling within this category. See Beard v. Banks, 542 U.S. 406, 417, 124 S.Ct. 2504, 159 L.Ed.2d 494 (2004).

And what does the Supreme Court use to determine whether a new rule qualifies as a new rule of substantive law? Our old friends the separation-of-powers doctrine and the principle of limited government powers, of course, since retroactivity doctrine is a constitutionally required aspect of habeas corpus, and habeas corpus, in turn, is grounded in the separation-of-powers doctrine and the principle of limited government powers. So as Supreme Court precedent convincingly demonstrates, the concern for keeping government powers in constitutional check drives the determination of whether a new rule qualifies as substantive, just as it does the jurisprudence of habeas corpus as a whole.

Indeed, Justice Harlan's Mackey concurrence defines substantive rules that warrant retroactivity as "those that place, as a matter of constitutional interpretation, certain kinds of primary, private individual conduct beyond the power of the criminal law-making authority to proscribe," 401 U.S. at 692, 91 S.Ct. 1160 (Harlan, J., concurring)—meaning rules that show that Congress has enacted legislation beyond its constitutionally authorized powers, in violation of the principle of limited government powers.

Nor has the Supreme Court limited the availability of retroactivity to claims involving only Congress's overstepping of its constitutional powers. In Bousley , the Supreme Court reasoned that the separation-of-powers doctrine rendered substantive—and thus retroactive—the new rule that it had announced in Bailey v. United States , 516 U.S. 137, 116 S.Ct. 501, 133 L.Ed.2d 472 (1995) —a case involving the Judiciary's overreaching of its constitutional powers. Bousley , 523 U.S. at 616-24, 118 S.Ct. 1604.

In Bailey , the Supreme Court construed 18 U.S.C. § 924(c)(1), which, at the time, imposed a prison term upon a person who "during and in relation to any ... drug trafficking crime ... uses or carries a firearm," to require evidence that the defendant actively employed the firearm during and in relation to the predicate crime. Bailey , 516 U.S. at 142-43, 116 S.Ct. 501. Previously, some courts had interpreted the provision to require evidence of only accessibility and proximity of a firearm during a drug-trafficking crime, not of active employment. Based on Bailey 's reading of § 924(c)(1), the Supreme Court concluded, courts exceeded their powers if they sentenced prisoners for an act that Congress did not make a crime under § 924(c)(1). As a result, the Supreme Court determined in Bousley that the Bailey rule had to be retroactively available.

In reaching this conclusion, the Supreme Court emphasized the important role that the separation-of-powers doctrine plays in habeas—and therefore retroactivity—analysis: "[U]nder our federal system it is only Congress, and not the courts, which can make conduct criminal.... Accordingly, it would be inconsistent with the doctrinal underpinnings of habeas review to preclude petitioner from relying on our decision in Bailey in support of his claim that his guilty plea was constitutionally invalid." Bousley , 523 U.S. at 620-21, 118 S.Ct. 1604 (citations omitted) (emphasis added). In other words, the separation-of-powers doctrine—and thus habeas corpus and the Suspension Clause—constitutionally requires new substantive rules, including those like the Bailey rule, which are statutory in nature, to be retroactively applicable and available to prisoners on collateral review.

And just last year, in Welch , the Supreme Court found that the separation of powers mandated the conclusion that the rule established in Johnson v. United States , ––– U.S. ––––, 135 S.Ct. 2551, 192 L.Ed.2d 569 (2015), is retroactive.¹⁵ To explain why, the Supreme Court hearkened back to its separation-of-powers reasoning in Bousley , observing that the separation-of-powers doctrine precludes a court from sentencing a person to more time in prison than Congress authorized, in the same way that it does not allow a court to imprison a person for an act that Congress did not validly criminalize:

15 The Supreme Court held in Johnson that the residual clause of the Armed Career Criminal Act of 1984, 18 U.S.C. § 924(e)(2)(B)(ii), is unconstitutionally vague and therefore void. See Johnson, 135 S.Ct. at 2555-63.

Bousley noted that the separation of powers prohibits a court from imposing criminal punishment beyond what Congress meant to enact.... But a court likewise is prohibited from imposing criminal punishment beyond what Congress in fact has enacted by a valid law. In either case a court lacks the power to exact a penalty that has not been authorized by any valid criminal statute.

Welch , 136 S.Ct. at 1268 (citation omitted).

The Supreme Court's pronouncements in Bousley and Welch illustrate Justice Harlan's observation 45 years ago that the "relevant frame of reference [for determining whether a new rule is retroactively applicable to cases on collateral review] ... is not the purpose of the new rule whose benefit the petitioner seeks, but instead the purposes for which the writ of habeas corpus is made available." Mackey , 401 U.S. at 682, 91 S.Ct. 1160 (Harlan, J., concurring). And we already know that the essential purpose of habeas corpus is to keep the government in check by zealously guarding the separation of powers and the principle of limited government. So it makes perfect sense that these same doctrines play a vital role in determining the retroactivity on collateral review of new rules of law.

True, the Supreme Court has also accounted for finality interests in criminal cases in its retroactivity framework. But the Supreme Court does not consider finality interests in a vacuum. Rather, the Court's retroactivity analysis "creates a balance between, first, the need for finality in criminal cases, and second, the countervailing imperative to ensure that criminal punishment is imposed only when authorized by law," Welch , 136 S.Ct. at 1266 —meaning only as permitted by the separation of powers and the principle of limited government powers.

Indeed, new substantive rules are retroactive for the very reason that "where the conviction or sentence in fact is not authorized by substantive law, ... finality interests are at their weakest. As Justice Harlan explained, '[t]here is little societal interest in permitting the criminal process to rest at a point where it ought properly never to repose.' " Id. (quoting Mackey , 401 U.S. at 693, 91 S.Ct. 1160 (Harlan, J., concurring)) (emphasis added).

Notably, Welch and Justice Harlan did not distinguish between illegal convictions and illegal sentences, emphasizing that in both cases, "finality interests are at their weakest." Id. So while I agree with Judge Jordan that equity "does not always draw clean lines, and the finality concerns embodied in § 2255(h) cannot be ignored," Jordan Op. at 1107–08, if, as Judge Jordan opines (and I agree), finality interests do not outweigh the interests in imposing punishment for a conviction when that conviction is not authorized by law, they are equally insufficient to outweigh the interests in imposing a sentence when that sentence is not authorized by law because it exceeds a congressionally imposed statutory maximum.

All of this—that is, the existence of retroactivity analysis—shows that, for purposes of understanding the meaning of "ineffective," timing is everything when it comes to having a "meaningful opportunity"—an opportunity that is not constitutionally deficient—to present claims that are ultimately governed by new rules of substantive law.

And as Bousley , which involved a new rule of statutory law, and Welch , which concerned a new rule of constitutional law, show, the separation-of-powers and limited-government-powers concerns are exactly the same for both retroactively applicable new rules of statutory law and retroactively applicable new rules of constitutional law. Compare Bousley , 523 U.S. at 620-21, 118 S.Ct. 1604 (statutory rule), with Welch , 136 S.Ct. at 1268 (constitutional rule). For this reason, new statutory and constitutional rules must be retroactively applicable on collateral review to the same extent. In both cases, a branch of the government has exceeded its constitutional powers. So on initial collateral review, a petitioner may make a claim based on the Supreme Court's new rule that did not exist during the petitioner's direct appeal, even though the petitioner had the opportunity to raise the same issue on direct appeal.

That's because not just any opportunity to raise an issue will do under habeas corpus jurisprudence; to comport with constitutional requirements, the opportunity to raise an issue must be meaningful. And an opportunity on direct review is not meaningful if a claim is denied and a new retroactively applicable rule subsequently establishes that the right not to be detained under the challenged provision in violation of the separation of powers or the principle of limited government always existed—regardless of whether that new rule is statutory or constitutional in nature.

Nor has the Supreme Court suggested that a different standard of retroactivity applies for second or successive claims than for initial claims on collateral review. Why would it? As we have discussed, the separation-of-powers doctrine and the principle of limited government powers are the animating reasons for why a new rule of substantive law must be retroactively applicable on collateral review. A sentence imposed in excess of the court's constitutional authority violates the separation of powers just as much if it is raised in a second or successive claim as it does if it is raised in an initial claim.

And an opportunity is not a meaningful one on an initial claim any more than it is on direct review if the initial claim is denied and a new retroactively applicable rule subsequently establishes that the right not to be detained under the detaining mechanism at issue in the case, in violation of the separation of powers and the principle of limited government powers, has always existed. Cf. 28 U.S.C. § 2255(h)(2) (permitting consideration of second or successive claims based on a retroactively applicable new rule of constitutional law). Because the Suspension Clause preserves habeas to protect against government action in excess of constitutional powers, the Suspension Clause requires consideration of second or successive claims that rely on a new retroactively applicable rule of law that was not available during direct appeal or earlier collateral proceedings.¹⁶ For it is here, "where the conviction or sentence in fact is not authorized by substantive law," that "finality interests are at their weakest." Welch , 136 S.Ct. at 1266.

16 For this reason, Bryant 's test, see 738 F.3d at 1274, which requires that binding precedent have foreclosed the prisoner's claim at the time of his first motion to vacate in order for the prisoner to access habeas through the saving clause, cannot be correct. Under Bryant, the prisoner whose claim is the one that establishes the precedent that squarely forecloses the claims of those who come after him has no meaningful opportunity to present his second or successive claim based on the new retroactively applicable rule of statutory law that allows those who follow to present their second or successive claims once the Supreme Court issues its new retroactively applicable rule of statutory law.

Of course, that is not to say that no finality interests are at play here, for limitations can and do exist on the habeas-corpus right to bring a claim based on a retroactively applicable new rule of substantive law. Habeas corpus entitles a petitioner to one meaningful opportunity to present his claim. So if, for example, a prisoner presents a claim based on a new retroactively applicable rule of law and loses, his habeas-corpus rights have been satisfied, and he may not continue to file new petitions raising the same claim, see 28 U.S.C. § 2244(a), without the Supreme Court's having issued an intervening new retroactively applicable rule.

C. Section 2255 does not otherwise permit consideration of second or successive claims based on a new rule of statutory law that reveals a violation of the separation of powers, so such claims must be cognizable under § 2255(e).

Now that we have established that habeas corpus requires consideration of a petitioner's claim based on a retroactively applicable new rule of substantive law—whether constitutional or statutory in nature and whether raised as an initial or second or successive collateral claim—we review whether § 2255, in fact, allows for consideration of such claims. Clearly, it does for claims based on retroactively applicable new constitutional rules. See 28 U.S.C. § 2255(h)(2). The express language of subsection (h)(2) provides that "[a] second or successive motion must be certified as provided in section 2244 by a panel of the appropriate court of appeals to contain—(2) a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable."

But nothing in § 2255 expressly allows consideration of second or successive claims raising a retroactively applicable new rule of statutory law. On the contrary, 28 U.S.C. § 2255(h), in conjunction with the provisions of § 2244 that it incorporates by reference, precludes consideration of claims relying on a retroactively applicable new rule of statutory law. Because the Suspension Clause requires consideration of these claims and yet § 2255 otherwise does not allow them, § 2255(e) must permit a prisoner to bring such claims in an application for writ of habeas corpus.¹⁷ And since McCarthan seeks to rely on a new retroactively applicable rule of statutory law that was not available when he filed his initial § 2255 petition, § 2255(e) authorizes consideration of his claim.¹⁸

17 A reader might wonder why Congress did not enumerate this type of claim along with the two types of claims listed in § 2255(h), particularly § 2255(h)(2). This question initially bothered me as well. So I extensively examined the legislative history for § 2255, but I could not find an explanation for why Congress enacted § 2255(h)(2). Based on my review of Supreme Court precedent, however, I believe the answer is that § 2255(h) was Congress's effort to ensure that constitutionally required second or successive claims not be precluded by AEDPA's amendments. But when Congress enacted § 2255(h) in 1996, the Supreme Court had not yet ruled that new statutory rules could be retroactive just like new constitutional rules could be. Instead, at that time, as far as claims based on retroactively applicable new rules were concerned, Congress likely understood the Constitution to require consideration of only those claims based on new substantive rules of constitutional law, as Justice Harlan's Mackey concurrence had suggested. See Mackey, 401 U.S. at 684, 91 S.Ct. 1160 (Harlan, J., concurring). So while I read § 2255(h) as Congress's attempt to ensure that § 2255 preserved habeas's constitutional scope, I read § 2255(e) in tandem as a failsafe mechanism that Congress continued to allow to exist because it recognized that it may have overlooked constitutionally required claims. Congress could have repealed § 2255(e) in 1996 if it intended § 2255(h) to render § 2255(e) superfluous, but it did not. To the extent that an argument might be made that Congress kept the saving clause for the separate reason that the clause was needed to provide relief where practical considerations arose, Congress could have amended the clause to expressly limit it to that situation, such as by explicitly referring to "practical considerations" or by removing the language "or ineffective." Again, it did not. And to the extent that some might note that habeas corpus did not always require what are now considered to be retroactively applicable new rules of statutory construction to be retroactively applicable, the Supreme Court has stated that "Felker, Swain, and Hayman stand for the proposition that the Suspension Clause does not resist innovation in the field of habeas corpus." Boumediene, 553 U.S. at 795, 128 S.Ct. 2229. As the Court has further explained, "[h]abeas is not a static, narrow, formalistic remedy; its scope has grown to achieve its grand purpose." Id. at 780, 128 S.Ct. 2229 (quoting Jones v. Cunningham, 371 U.S. 236, 243, 83 S.Ct. 373, 9 L.Ed.2d 285 (1963) (internal quotation marks omitted)). So the fact that Congress may have viewed the scope of habeas narrowly in 1996 is no answer to the Supreme Court's current explanation of the Suspension Clause's constitutional scope.

18 Of course, to the extent that McCarthan procedurally defaulted the claim and the government raises this as an affirmative defense, McCarthan would have to show "cause and prejudice" for the default or "actual innocence" in order for his claim to be considered on the merits. See Bousley, 523 U.S. at 621-22, 118 S.Ct. 1604.

Because the district court did not consider McCarthan's claim on the merits in the first instance, I would reverse the district court's dismissal of McCarthan's claim and remand for consideration of the merits.

II.

The Majority's criticism of my proposed interpretation of the saving clause focuses on three things: (1) my reliance on Boumediene , 553 U.S. 723, 128 S.Ct. 2229 ; (2) my conclusion that retroactivity doctrine is constitutionally required; and (3) the alleged mooting of my theory by Felker v. Turpin , 518 U.S. 651, 116 S.Ct. 2333, 135 L.Ed.2d 827 (1996). The Majority is mistaken on all three counts.

A. The Majority's attack on my reliance on Boumedienedoes not withstand scrutiny.

The Majority takes issue with my reliance on Boumediene because that case "addressed the scope of habeas corpus for executive detainees 'where no trial has been held' and distinguished decisions like Felker , in which a prisoner sought relief from a judgment imposed in a 'fair, adversary proceeding.' " Maj. Op. at 1094–95 (quoting Boumediene , 553 U.S. at 732, 774, 782, 128 S.Ct. 2229 ). On its face, this might seem like a good way of distinguishing Boumediene —until we look at how this dissent actually uses Boumediene .

First, I rely on Boumediene primarily for citation of principles universally applicable to habeas corpus and the Suspension Clause—regardless of the posture of the litigating prisoner. For example, I cite Boumediene for its interpretation of Swain and Hayman as standing for the proposition that failure to interpret the saving clause as a constitutional failsafe would raise "serious question[s] about the constitutionality of [§ 2255 ]." Swain and Hayman involved prisoners who had already gone through a presumably "fair, adversary proceeding" and had been sentenced, and the fact that Boumediene involved Guantanamo prisoners instead of sentenced federal prisoners does not somehow render the lessons of Swain and Hayman any less correct or applicable.

Similarly, to the extent that the Majority's criticism of my reliance on Boumediene is intended to apply to my citations of Boumediene for the principle that concerns of separation of powers and limited government powers animate habeas corpus, and therefore the Suspension Clause, the Majority likewise provides no explanation why the fact that Boumediene involved Guantanamo prisoners somehow makes that general principle less applicable in the case of sentenced federal prisoners. After all, there is only one Suspension Clause, and it applies to executive-branch prisoners and federally sentenced prisoners alike.

B. The Majority mistakenly views retroactivity doctrine as an exception to procedural barriers.

And even if we did not consider Boumediene , the Majority's criticism does not properly account for Supreme Court jurisprudence on retroactivity—an area that directly involves and applies to sentenced federal prisoners. As I have noted, the Supreme Court in retroactivity doctrine has relied on the very same separation-of-powers and limited-government-powers concerns as it has in explaining in Boumediene the driving forces behind habeas and the Suspension Clause. See supra at 1137–38 ("[U]nder our federal system it is only Congress, and not the courts, which can make conduct criminal.... Accordingly, it would be inconsistent with the doctrinal underpinnings of habeas review to preclude petitioner from relying on our decision in Bailey in support of his claim that his guilty plea was constitutionally invalid." (quoting Bousley , 523 U.S. at 620-21, 118 S.Ct. 1604 (citations and quotation marks omitted) (emphasis added))); see id. at 145, 116 S.Ct. 501 (" Bousley noted that the separation of powers prohibits a court from imposing criminal punishment beyond what Congress meant to enact.... But a court likewise is prohibited from imposing criminal punishment beyond what Congress in fact has enacted by a valid law. In either case a court lacks the power to exact a penalty that has not been authorized by any valid criminal statute." (quoting Welch , 136 S.Ct. at 1268 (citation and quotation marks omitted))).

That is no accident. Retroactivity doctrine is based on "the doctrinal underpinnings of habeas review," Bousley , 523 U.S. at 621, 118 S.Ct. 1604 —meaning the Suspension Clause's concerns for separation of powers and the principle of limited government powers—because it is required by "the doctrinal underpinnings of habeas review" and therefore by the Suspension Clause. This fact seems to escape the Majority, so it incorrectly suggests that Congress could do away altogether with collateral review of claims that rely on retroactively applicable rules if it so desired. See Maj. Op. at 1095 ("Retroactivity means that a court is no longer barred from applying a new rule on collateral review, not that a court must create a vehicle for collateral review because there is a new rule."). But that is not an option because the "doctrinal underpinnings of habeas review"—and therefore the Suspension Clause—constitutionally require retroactivity doctrine and therefore a procedural vehicle for collateral review of claims that rely on retroactively applicable new rules.

So even without considering Boumediene , the Suspension Clause requires consideration of second or successive claims that involve retroactively applicable new rules of statutory law because they expose detention in violation of the "doctrinal underpinnings of habeas review"—that is, detention in violation of the separation of powers and the principle of limited government powers.

The Majority has no sufficient answer to this. Instead, it responds by confusing two distinct concepts: retroactivity doctrine and procedural barriers. See id . First, the Majority states, "When the Supreme Court makes a right retroactively available on collateral review, it does not mean that a prisoner is constitutionally entitled to have a court review a violation of that right on the merits." Id. I agree. But a prisoner is constitutionally entitled to a meaningful opportunity to have a court review a violation of that right on the merits.

The Majority does not appreciate the difference between a meaningful opportunity to have a claim heard and an unqualified constitutional right to have a claim heard, regardless of the prior availability of meaningful opportunities to have done so. So it next invokes the existence of procedural barriers such as the procedural-default rule and the statute of limitations as evidence supporting its incorrect premise. Id.

To be sure, procedural barriers like procedural default and the statute of limitations may constitutionally preclude a prisoner from bringing his claim on a new retroactively applicable rule. But that has nothing to do with why retroactivity doctrine is or is not constitutionally required. Nor does it have anything to do with whether a prisoner has a meaningful opportunity to bring his claim if he procedurally defaults it or files it after the statute of limitations has expired.

Procedural barriers like procedural default and the statute of limitations can constitutionally limit access to collateral review through retroactivity doctrine because even when they act to bar a claim, they do not bar a meaningfulopportunity to present the claim. In other words, it is possible to obtain collateral review, so long as the petitioner complies with these procedural rules.

But that is not the case with a second or successive claim that rests on a new retroactively applicable rule of statutory law. At every stage of the proceedings through initial collateral review, a petitioner may raise the argument on which the new rule is based (thereby not procedurally defaulting) and may file a second or successive claim within a year of the Supreme Court's announcement of a retroactively applicable new rule of statutory law. But in the absence of the saving clause, § 2255 provides no opportunity to present the claim if the Supreme Court does not recognize, until after the initial collateral claim has been disposed of, that the right has always existed. And since the doctrinal underpinnings of habeas constitutionally require one meaningful opportunity to present a claim based on a retroactively applicable rule of statutory law, the Majority's construction of § 2255 not to provide such an opportunity amounts to a suspension of the writ.

C. Felker does not solve the Majority's problem .

In response to this problem, the Majority invokes Felker . Maj. Op. at 1095. The Majority reasons that its own "interpretation of the saving clause cannot suspend the writ because the Original Writ in the Supreme Court remains available, habeas corpus at common law did not apply to prisoners sentenced by a court of competent jurisdiction, and the decision of the Supreme Court in Felker ... upheld a bar on successive motions against constitutional challenge." Maj. Op. at 1093–94. None of these arguments saves the Majority's interpretation of the saving clause from constitutional peril.

Beginning with the Majority's third argument first, the Majority asserts that the Supreme Court's opinion in Felker supports the conclusion that the limitations that § 2255(h) places on second or successive claims are absolute and constitutional. This argument is based on the premise that Felker held that the second or successive restrictions for state prisoners seeking federal habeas relief, codified at 28 U.S.C. § 2244(b),¹⁹ do not violate the Suspension Clause. See Felker , 518 U.S. at 664, 116 S.Ct. 2333. Because those restrictions are significantly similar to the second or successive restrictions in § 2255(h), and because the limitations on successive state petitions do not contain a saving clause like § 2255 does, the argument goes, the § 2255(h) restrictions must be constitutional if the § 2244(b) restrictions are constitutional. See Maj. Op. at 1095.

19 Section 2244(b) provides,

(b)(1) A claim presented in a second or successive habeas corpus application under section 2254 that was presented in a prior application shall be dismissed.

(2) A claim presented in a second or successive habeas corpus application under section 2254 that was not presented in a prior application shall be dismissed unless—

(A) the applicant shows that the claim relies on a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable; or

(B)(i) the factual predicate for the claim could not have been discovered previously through the exercise of due diligence; and

(ii) the facts underlying the claim, if proven and viewed in light of the evidence as a whole, would be sufficient to establish by clear and convincing evidence that, but for constitutional error, no reasonable factfinder would have found the applicant guilty of the underlying offense.

The logic is superficially appealing, but it fails to take into account a fundamental difference between federal prisoners' collateral claims and state prisoners' habeas claims: separation-of-powers and limited-government-powers concerns can constitutionally require habeas relief for federal prisoners, while these doctrines are irrelevant to determining the availability of habeas relief for state prisoners. Federal courts adjudicating federal prisoners' claims like McCarthan's enforce federal separation-of-powers values by correcting convictions or sentences where an Article III court exceeded its congressionally authorized powers.

That aspect of a federal collateral proceeding like McCarthan's is not present in a federal-court adjudication of a state prisoner's habeas petition. A state court cannot act in violation of the federal separation of powers because the state court is not a part of the federal Judiciary. So not surprisingly, § 2254 does not contain a saving clause, since unlike with habeas review of federal cases, habeas review of state cases does not raise separation-of-powers or limited-government-powers concerns. As a result, the Court in Felker had no reason to and did not consider the separation of powers in its analysis because it was irrelevant in that case. Here, in contrast, the separation-of-powers doctrine drives the analysis of the saving clause's meaning.

The Majority responds to this important difference between federal and state habeas by calling it " 'interpretive jiggery-pokery' " and suggesting that I "ignore[ ]" the fact that the principle of limited government powers also animates habeas corpus. Maj. Op. at 1095 (quoting King v. Burwell , ––– U.S. ––––, 135 S.Ct. 2480, 2500, 192 L.Ed.2d 483 (2015) (Scalia, J., dissenting)). Setting aside the irony in the Majority's charge, the Majority misses the point: the principle of limited government powers in the context of habeas is the principle that the federal government may not exceed the powers granted to it by the Constitution. Neither this concept nor the separation-of-powers doctrine is in play when a federal court grants a state prisoner's habeas petition. In that context, the state's adjudication of a claim yields to federal interpretation of the governing law under principles of federalism and the supremacy of federal law—concepts distinct from the separation-of-powers doctrine and the principle of limited government that drive habeas for federal prisoners.

The Majority also suggests that we need not consider the Suspension Clause in interpreting the saving clause because a would-be petitioner like McCarthan has an alternate route for obtaining the relief he seeks: he may petition for an "Original Writ" from the Supreme Court. In support, the Majority cites Judge Pryor's concurrence in Samak , which, in turn, cites Felker . See Maj. Op. at 1094.

But a careful reading of Felker dispels this notion that the existence of the Original Writ allows Congress to preclude relief for second or successive claims required under the Suspension Clause to be permitted. To be sure, Felker relied on the availability of the Supreme Court's original jurisdiction to strike down a challenge to § 2244(b) under the Exceptions Clause.²⁰ Notably, though, it chose not to rely on the availability of the Court's original jurisdiction to strike down a challenge to § 2244(b) under the Suspension Clause.

20 The Exceptions Clause provides, in relevant part, "In all the other Cases ... the supreme Court shall have appellate Jurisdiction, both as to Law and Fact, with such Exceptions, and under such Regulations as the Congress shall make." U.S. Const. art. III, § 2, cl. 2.

In Felker , the Court held that § 2244(b)'s stripping of the Supreme Court's appellate jurisdiction to review a court of appeals's denial of a state prisoner's motion for leave to file a second habeas petition in the district court did not affect the Supreme Court's original jurisdiction to entertain original habeas petitions. See Felker , 518 U.S. at 658-61, 116 S.Ct. 2333. This conclusion consequently "obviate[d]" the need to address the constitutionality of Congress's jurisdiction-stripping under the Exceptions Clause; since the Supreme Court retained its jurisdiction to entertain an original petition, the Court reasoned, "no plausible argument" could exist that § 2244(b) violated the Exceptions Clause. Id. at 661-62, 116 S.Ct. 2333.

Then the Court turned for the first time to the challenge to § 2244(b) under the Suspension Clause. Significantly, the Court did not rule that those restrictions did not violate the Suspension Clause because the "Original Writ" was still available in the Supreme Court for any of the number of claims precluded by the second-or-successive bar. In fact, the Court in its analysis of the Suspension Clause issue did not even consider the Supreme Court's original jurisdiction to hear habeas petitions. Instead, the Court held that the § 2244(b) restrictions passed muster under the Suspension Clause because they substantially mirrored common-law habeas rules. See id. at 664, 116 S.Ct. 2333. Felker thus requires us to ask not whether the narrowest avenue for relief, such as the Original Writ, still exists, but rather whether the statutory provision at issue adequately substitutes for common-law habeas rules.When it comes to second or successive claims based on retroactively applicable new rules of statutory law, the availability of the Original Writ does not adequately substitute for common-law habeas rules for three reasons.

First, it is not even clear that the Supreme Court would have original jurisdiction to entertain a claim that is expressly precluded by § 2255(h). If, as the Majority and Chief Judge Carnes's concurrence have effectively argued, see Maj. Op. at 1093–94; E. Carnes Op. at 1100, only a prisoner whose second or successive claim complies with § 2255(h)'s bars is "authorized" to proceed under § 2255 and therefore under the saving clause, § 2255(h)'s bars would then also arguably restrict the Supreme Court. Cf. Felker , 518 U.S. at 662-63, 116 S.Ct. 2333 (pondering the applicability of §§ 2244(b)(1) and (2)'s unqualified restrictions on the filing of second or successive state habeas claims, to the Supreme Court's review).

Second, even if subsection (h)'s bars do not apply to the Supreme Court, the Original Writ does not provide a federal prisoner with a "meaningful opportunity" to test the legality of his detention—despite the fact that common-law habeas rules demand such a meaningful opportunity. Under Supreme Court Rule 20.4(a), the writ is "rarely granted" and only when "exceptional circumstances warrant the exercise of the Court's discretionary powers." So consideration of claims based on a retroactively applicable new rule of statutory law would not occur as a matter of course. As a result, while the Supreme Court's original jurisdiction might provide an opportunity, it would not be a meaningful one.

That leads to the third point. Even if the Supreme Court had upheld § 2244(b) against the Suspension Clause challenge on the basis of the availability of the Original Writ—which it did not—the procedure at issue in Felker was materially different—and far more conducive to Original Writ review—than that at issue under the saving clause. In Felker , the Court relied on its original jurisdiction to safeguard against individual incorrect decisions that an appellate court might make in denying the availability of habeas relief on the merits of a given case. We would hope these cases would be relatively few, but in any event, the § 2244(b) procedure does not on its face deny relief to an entire class of prisoners who we know are entitled to relief.

That, however, would not be the situation if prisoners with second or successive claims based on retroactively applicable new rules of statutory law all had to use the Original Writ to obtain relief to which the Suspension Clause entitles them. Then the Original Writ would have to serve as a regular mechanism for an entire class of prisoners—not just an individual prisoner here or there who was mistakenly denied the opportunity for habeas relief by an appellate court—to obtain relief. Imagine, for example, the Supreme Court's having to entertain in the first instance all second or successive claims based on a retroactively applicable new rule of statutory law! That cannot be what is contemplated by the Court's "rare[ ]" and "exceptional" use of its Original Writ jurisdiction. See Felker , 518 U.S. at 665, 116 S.Ct. 2333 (quoting Sup. Ct. R. 20.4(a)).

Nor does the Majority's reliance on Article III, § 1, of the Constitution adequately address the problem. See Maj. Op. at 1094. Invoking this constitutional provision, the Majority reasons that under its interpretation of the saving clause, the existence of the Original Writ must suffice to maintain the constitutionality of § 2255 under the Suspension Clause "because the Constitution does not even require Congress to create inferior courts." Id. In other words, in the Majority's view, Congress could enact legislation that completely abolished habeas corpus, as long as it preserved the Original Writ.

I respectfully disagree. Such a system would be entirely unworkable in today's world. Even setting aside the fact that abolition of habeas-corpus review outside the Supreme Court would require more than "exceptional" or "rare" review by Original Writ, it is difficult to conceive of how the nine Justices would be able to timely process in the first and last instance all habeas petitions from around the country, in addition to maintaining the rest of their important workload. And a system that systematically deprives prisoners of timely habeas review is a system that provides no meaningful habeas review. Unquestionably, the lack of a meaningful habeas review system would violate the Suspension Clause.

So putting the Majority's argument in the best light, the question is one of degree. I respectfully submit that using the Original Writ as a regular and only processing mechanism for an entire class of second or successive claims that are constitutionally required to be considered amounts to not providing a meaningful habeas-review system for those claims. It bears repeating that the claims we are talking about are required by the Suspension Clause to receive consideration.

As for the Majority's argument that habeas corpus at common law did not apply to prisoners sentenced by a court of competent jurisdiction, Maj. Op. at 1094–95, the Supreme Court itself has noted that " Felker , Swain , and Hayman stand for the proposition that the Suspension Clause does not resist innovation in the field of habeas corpus." Boumediene , 553 U.S. at 795, 128 S.Ct. 2229. As the Court has further explained, "[h]abeas is not a static, narrow, formalistic remedy; its scope has grown to achieve its grand purpose." Id. at 780, 128 S.Ct. 2229 (quoting Jones , 371 U.S. at 243, 83 S.Ct. 373 (internal quotation marks omitted)). And since retroactivity doctrine is constitutionally required under current habeas and Suspension Clause jurisprudence, the Majority's argument in this regard cannot provide a basis for denying a meaningful opportunity for collateral review to prisoners whose second or successive claims rest on a new, retroactively applicable rule of substantive law.

III.

Turning now to the Majority's interpretation of the saving clause, it suffers from three fatal flaws. First, it defies the statutory text it purports to respect. Second, it shortchanges Supreme Court jurisprudence on the Suspension Clause and habeas corpus. And third, it inexplicably limits the meanings of the general definitions it cites to shoehorn them into supporting the Majority's incorrect construction of the saving clause.

A. The Majority's interpretation of the saving clause contradicts the text of § 2255 .

Though I do not question my colleagues' earnestness in their proposed construction of the saving clause, the Majority opinion's approach to deciphering the text of § 2255, disappointingly, is like something right out of Alice's Adventures in Wonderland .²¹ True, the Majority opinion repeatedly claims that its analysis does nothing more than apply the plain meaning of § 2255's text. See, e.g. , Maj. Op. at 1080, 1082, 1084–93, 1095–97, 1099. But saying so—no matter how many times—doesn't make it so. In fact, the Majority opinion's interpretation of the statutory text clashes in significant ways with what the text actually says.

21 Lewis Carroll, Alice's Adventures in Wonderland (BookVirtual Digital Ed. v.1.2 2000), available at https://www.adobe.com/be_en/active-use/pdf/Alice_in_Wonderland.pdf (last accessed Jan. 3, 2017).

For starters, the Majority opinion seems most confused when it asserts that execution-of-sentence claims may be brought under the saving clause. In support of its theory, the Majority opinion reasons that § 2255 is " 'inadequate or ineffective to test' a prisoner's claim about the execution of his sentence because that claim is not cognizable under section 2255(a) ." Maj. Op. at 1088 (emphasis in original deleted; emphasis added).

But that's exactly why execution-of-sentence claims cannot be brought under the saving clause—because those claims are not cognizable under § 2255. Proceeding on a type of claim that § 2255 actually authorizes is an absolute prerequisite for a habeas petition to be considered under the saving clause. See supra at 1124–29; see also 28 U.S.C. § 2255(e) (the habeas petition of a prisoner "authorized to apply for relief by [§ 2255 ] motion ... shall not be entertained ... unless it also appears that the remedy by [§ 2255 ] motion is inadequate or ineffective to test the legality of his detention") (emphasis added). The universe of claims that § 2255"authorize[s]," however, does not include execution-of-sentence claims—even by the Majority's admission. Nor would it make sense for execution-of-sentence claims to be considered under the saving clause because they are authorized under § 2241, without regard to § 2255. See Antonelli , 542 F.3d at 1352.

So bringing an execution-of-sentence claim under § 2255's saving clause would be a lot like putting scuba gear on a swordfish so it could breathe underwater; neither the saving clause nor scuba gear is designed, is necessary, or works for the asserted purpose (or, as the Mock Turtle in Wonderland might say, "for the asserted porpoise"). See Carroll, supra , at 154-55.

The Majority responds to this problem by saying that the defect in its theory "proves nothing about whether a prisoner with a claim based on a change in caselaw or a prisoner with a claim based on actual innocence satisfies the saving clause." Maj. Op. at 1093–94. True. It doesn't.

But it's not intended to. The first section of this dissent already explains why a claim based on a new retroactively applicable rule of statutory interpretation—not just any "change in caselaw," as the Majority incorrectly characterizes my dissent as arguing—must be considered under the saving clause. I write about the incorrectness of the Majority's argument solely in an effort to prevent our Circuit from supplanting eighteen years of precedent with another interpretation of the law that cannot possibly be correct. After all, "if you drink ... from a bottle marked 'poison,' it is almost certain to disagree with you, sooner or later." Carroll, supra , at 10.

The Majority's textual confusion does not end with its incorrect insistence that execution-of-sentence claims may be brought under the saving clause. Rather, the Majority's interpretation just gets "[c]uriouser and curiouser!" Carroll, supra , at 15. The Majority also asserts that prisoners " '[kept] in custody' without a criminal sentence" may have their claims heard through the saving clause. Maj. Op. at 1089.

But that is precisely the opposite of what the saving clause's unambiguous language permits. Only prisoners who have already been sentenced by a federal court are eligible to take advantage of the saving clause—and only they need to do so to access habeas relief because prisoners who have not yet been sentenced are not subject to § 2255's strictures and may proceed directly under § 2241. See supra at 1124–29. That's because the saving clause allows consideration of claims presented by only those prisoners "authorized to apply for relief by [§ 2255 ] motion." 28 U.S.C. § 2255(e). Section 2255, in turn, authorizes the claims of only those prisoners "under sentence of a court established by Act of Congress." Id. § 2255(a) (emphasis added).

Not only that, but § 2255 authorizes solely those claims challenging sentences. See id. Only in Wonderland would it make sense for an unsentenced prisoner to bring a claim challenging his sentence. Cf. Carroll, supra , at 187 ("Sentence first—verdict afterwards.").

But the Majority goes even deeper down the rabbit hole, asserting next that the saving clause does not allow for consideration of second or successive claims that do not satisfy the requirements of subsection (h)'s bar on second or successive § 2255 motions. See Maj. Op. at 1089–91; see also E. Carnes Op. at 1100. The Majority does not look to the language of the saving clause in divining this supposed rule; instead, it relies incorrectly on the general/specific canon of construction and again does not account for the saving clause's failsafe function or apply the actual language of § 2255.

Invoking the general/specific canon, the Majority contends, "The specific language of section 2255(h), enacted nearly 50 years after the saving clause, limits the reach of the saving clause." Maj. Op. at 1090–91. First, this argument implicitly concedes that the language of the saving clause does not require a prisoner's claim to satisfy subsection (h) in order for a court to consider it under the saving clause. For if it did, the Majority would not need to invoke this canon.

Second, to the extent that the general/specific canon even applies to the analysis of § 2255, in invoking the general/specific canon of construction, the Majority gets things backwards; it wrongly characterizes the saving clause as the general provision and subsection (h) as the more specific provision. The general/specific canon states that "[i]f there is a conflict between a general provision and a specific provision, the specific provision prevails." Scalia & Garner, supra , at 183. This canon relies on the principle that "the two provisions are not in conflict, but can exist in harmony." Id. at 185. As Justice Scalia and Bryan Garner have explained, "The specific provision does not negate the general one entirely, but only in its application to the situation that the specific provision covers." Id.

Applying the general/specific canon can be challenging because it is not always easy to ascertain which provision lays the general rule and which the specific. Id. at 187; see also id. at 188 (quoting Radzanower v. Touche Ross & Co. , 426 U.S. 148, 159, 96 S.Ct. 1989, 48 L.Ed.2d 540 (1976) (Stevens, J., dissenting)). And timing of enactment does not necessarily reveal the answer. Id. at 187. Here, the bars contained in subsection (h), though enacted later in time, establish the general rule that courts may consider only those second or successive claims that satisfy the conditions of subsection (h). The saving clause provides the exception: if failure to consider a second or successive claim that does not satisfy the criteria of subsection (h) would result in a violation of the Suspension Clause, the saving clause requires consideration of that claim.

The general and specific labels make no sense if applied in the other direction. It would be "uncommon nonsense," Carroll, supra , at 158, indeed, to allow the unconstitutional provision (subsection (h), in the absence of subsection (e)'s failsafe mechanism) to trump the provision whose job it is to ensure subsection (h)'s and the rest of § 2255's constitutionality. So subsection (h)'s bar cannot preclude consideration of second or successive claims that the Suspension Clause—and therefore the saving clause—requires.

Each of the three errors discussed above—the insistence that execution-of-sentence claims may be considered under the saving clause; the suggestion that unsentenced prisoners' claims may be considered under the saving clause; and the notion that subsection (h), which purports to bar second or successive claims that are constitutionally required to be considered, takes priority over the saving clause's constitutional-failsafe mechanism—are examples of "needlessly rendering provisions in conflict [though] they can be interpreted harmoniously." See Maj. Op. at 1090 (quoting Scalia & Garner, supra , at 180) (quotation marks omitted). As the Majority itself notes, though, "[t]here can be no justification" for such an analysis. Id. (quoting Scalia & Garner, supra , at 180) (quotation marks omitted).

Finally, the Majority expresses concern that construing the saving clause to allow consideration of second or successive claims that may not be entertained under subsection (h) allows saving-clause claims to avoid subsection (f)'s statute of limitations. Maj. Op. at 1091. But that is doubly wrong, even if, as the Majority suggests, subsection (f)'s statute of limitations is incorporated into subsection (e), see Maj. Op. at 1093–94 (arguing that subsections (f) and (h) "authorize" a prisoner to apply for relief under § 2255 and therefore are incorporated into the saving clause's requirement that a prisoner be "authorized" to apply for relief under § 2255 in order to have his second or successive claim considered under the saving clause).

First, under subsection (f)(3), courts may consider claims filed within one year of "the date on which the right asserted was initially recognized by the Supreme Court, if that right has been newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review." 28 U.S.C. § 2255(f)(3). Among other functions, this provision allows courts to entertain (h)(2) claims—"contain[ing] ... a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court that was previously unavailable," id. § 2255(h)(2) —even if more than a year of untolled time has passed since the prisoner's conviction became final.

If claims satisfying subsection (h)(2)'s criteria were not viewed as asserting a "right [that] has been newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review," id. § 2255(f)(3), they could not be considered under § 2255's statute of limitations if they were brought more than a year after the prisoner's conviction became final because subsection (f)(3) would not apply to them, and subsection (f)(3) would lack a function. This interpretation would violate the surplusage canon. See Scalia & Garner, supra , at 174-79; see also Reiter v. Sonotone Corp ., 442 U.S. 330, 339, 99 S.Ct. 2326, 60 L.Ed.2d 931 (1979) ("In construing a statute we are obliged to give effect, if possible, to every word Congress used.").

But that's not how courts apply § 2255's statute of limitations to second or successive claims that rely on a retroactively applicable new rule of constitutional law. Necessarily then, second or successive claims that "contain ... a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable," 28 U.S.C. § 2255(h)(2), must also assert a "right [that] has been newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review," id. § 2255(f)(3).

That right is the right protected by the Suspension Clause not to be detained in violation of the separation of powers or the principle of limited government powers, through the impermissible interpretation of the statute or other device used to detain the prisoner, that has since been invalidated by the Supreme Court's retroactively applicable new rule of substantive law—whether constitutional or statutory. And since subsection (f)(3) imposes a one-year statute of limitations on all claims involving retroactively applicable rights—without respect to whether they are based on a new rule of statutory or constitutional law—if it is incorporated into subsection (e), it necessarily imposes a one-year statute of limitations on second or successive claims that raise a right that depends on a new rule of statutory law made retroactively applicable by the Supreme Court to cases on collateral review. As a matter of fact, then, accounting for the Majority's view that subsection (e) incorporates subsection (f)'s statute of limitations, second or successive claims that depend on new retroactively applicable rules of statutory law do not evade § 2255's statute of limitations.²²

22 The upshot of this fact is that the Majority's contention that subsections (f) and (h) "authorize" prisoners to apply for relief by § 2255 motion, see Maj. Op. at 1093–94, has no bearing on the ultimate construction of the saving clause to require consideration of second or successive claims that rely on a new retroactively applicable rule of statutory law. Even if subsections (f) and (h) "authorize" prisoners to apply for relief by § 2255 motion, that means that subsection (f) necessarily imposes a one-year statute of limitations on second or successive claims that rely on a new retroactively applicable rule of statutory law, which does not begin to run until the Supreme Court recognizes the right by announcing the new rule, and subsection (h) must give way to the saving clause to the extent that its failure to do so would cause § 2255 to violate the Suspension Clause.

Second, even if second or successive claims based on a "right [that] has been newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review," id. § 2255(f)(3), do avoid the statute of limitations,²³ that is not sufficient grounds to disregard the saving clause's constitutional-failsafe function and interpret § 2255 to prohibit claims that are constitutionally required. If Congress wants to impose a statute of limitations on saving-clause claims, nothing stops it from doing so.²⁴ The same is true of requiring prisoners to seek a certificate of appealability to bring claims based on a new retroactively applicable rule of statutory law. B. The Majority's Interpretation of the Saving Clause Ignores Critical Supreme Court Precedent.

23 The Majority inaccurately characterizes this dissent as taking the position that any prisoner who fails any procedural bar in section 2255 can petition for a writ of habeas corpus. Maj. Op. at 32-33. That has never been my argument. Rather, I contend that a prisoner whose second or successive sentencing claim is otherwise barred by § 2255 from being brought may have his claim considered under the saving clause only if failure to consider the claim would render § 2255 constitutionally deficient, as in the case of second or successive claims based on a retroactively applicable new rule of statutory law.

24 Indeed, that's exactly what Congress did with respect to the claims identified at § 2255(f) when it passed AEDPA. Before AEDPA was enacted, habeas doctrine allowed courts to consider petitions "filed after even extraordinary delays." Day v. McDonough, 547 U.S. 198, 215, 126 S.Ct. 1675, 164 L.Ed.2d 376 (2006) (Scalia, J., dissenting) (citing cases where petitions were entertained 40 years, 36 years, and 24 years after the filing prisoner was sentenced).

To support its interpretation of the saving clause, the Majority's opinion purports to rely primarily on dictionaries, Judge William Pryor's concurrence in Samak v. Warden, FCC Coleman-Medium , 766 F.3d 1271, 1275-95 (11th Cir. 2014) (W. Pryor, J., concurring), and the Tenth Circuit's decision in Prost v. Anderson , 636 F.3d 578 (10th Cir. 2011) (on which Judge Pryor's Samak concurrence, in turn, relies).

Prost , however, specifically declined to consider "whether, when, and how the application of § 2255(h)'s limits on second or successive motions might (ever) raise a serious constitutional question."²⁵ Id. at 594 ; see also id. at 583 n.4 (acknowledging that the Suspension Clause may require "some avenue of collateral attack"). So Prost 's analysis does not account for—and does not purport to account for—the Suspension Clause and its attendant constitutional considerations. Yet Prost expressly recognized that to the extent that § 2255's bans on second or successive motions violated the Constitution, § 2255 could not be upheld. See id. at 586 n.6 ("[Courts] [h]aving created our own (if different) hierarchy of innocence claims, it's hard to say Congress wasn't entitled to enact its own—unless of course its particular hierarchy in some way can be said to violate the Constitution.").

25 Though the Majority claims to follow Prost 's interpretation of the saving clause, see Maj. Op. at 1080, Prost also does not adopt the Majority's position that the saving clause authorizes execution-of-sentence and pretrial-detention claims. That construction appears to have originated as purely an invention of Judge Pryor's. See Samak, 766 F.3d at 1291 (W. Pryor, J., concurring).

We are in a different posture than the Tenth Circuit. Today we abrogate nearly 20 years of Circuit precedent on our interpretation of the saving clause and replace it with an entirely different framework of understanding. When we undertake this type of drastic change in our jurisprudence, we have the deepest of responsibilities to try to ensure that our new solution is correct—or at least that it has accounted for constitutional concerns.

But the Majority gives short shrift to the Suspension Clause and barely mentions retroactivity doctrine or the importance in habeas jurisprudence of the separation-of-powers doctrine and the principle of limited government powers. Having a discussion about the saving clause without delving into all of these constitutional concepts is a lot like trying to play football without a ball: it can't be done correctly.

The point of the saving clause is to save § 2255 from any potential unconstitutionality. See supra at 1132–34; see also Boumediene , 553 U.S. at 776, 128 S.Ct. 2229. If § 2255 did not permit consideration of claims that are constitutionally required, it would be unconstitutional. See supra at 1132–34. So addressing the Constitution's—and in particular, the Suspension Clause's—requirements is critical to the correct interpretation of the saving clause. And since the separation-of-powers doctrine and the principle of limited government powers drive the Suspension Clause's protection of habeas claims, we must consider the role of those doctrines as well when we construe the saving clause. Finally, because these same concerns require retroactivity of new rules of both constitutional and statutory law, we must also account for retroactivity in our analysis.

But the Majority's analysis dismisses all of these concerns without adequate reasoning. The Majority's reliance on Felker for its Original-Writ answer to the Suspension Clause problem fails because Felker does not hold that Suspension Clause problems may be resolved by the existence of original jurisdiction. See supra at 1143–48. Nor does the Majority address the importance of the separation-of-powers doctrine and the principle of limited government powers in habeas and Suspension Clause jurisprudence. And while the Majority Opinion mentions "retroactivity," it does not acknowledge that retroactivity exists because it is constitutionally required. The Majority likewise neglects to give any consideration to the constitutional reasons requiring retroactivity. See Maj. Op. at 1095; see supra at 1141–43.

Instead, the Majority opines that determining whether a new rule of law "applies retroactively on collateral review can be a difficult and controversial task" and then employs that opinion as a reason to support its interpretation of the saving clause because the Majority's construction conveniently does not require courts to determine whether a new rule applies retroactively. Maj. Op. at 1098–99. But courts' convenience cannot excuse failure to comply with constitutional requirements. And the Majority does not account for the constitutional basis for retroactivity doctrine at all in arriving at its construction of the saving clause.

Because the Majority does not adequately consider the Suspension Clause, the doctrine of separation of powers and the principle of limited government powers, and retroactivity doctrine, the Majority does not recognize that the saving clause requires consideration of second or successive claims that rely on a new retroactively applicable rule of statutory law.²⁶ As a result, the Majority's interpretation of the saving clause renders § 2255 unconstitutional.

26 Again, the Majority misrepresents my dissent when it asserts in response, "[T]he writ has not been suspended whenever a prisoner cannot file a successive collateral attack." Maj. Op. at 1095. That has never been my argument. Instead, I contend that a procedural mechanism that fails to allow for consideration of second or successive claims that rely on a new retroactively applicable rule of statutory law violates the Suspension Clause.

C. The Majority's interpretation relies on artificially limited definitions, contorted to fit the Majority's desired construction of the saving clause.

The Majority identifies and defines four terms from the saving clause: "remedy," "to test," "inadequate or ineffective," and "detention." Maj. Op. at 1085. None of these terms requires the construction that the Majority devises today. And none precludes the interpretation of the saving clause to permit second or successive claims based on a retroactively applicable new rule of statutory law.

The Majority begins by defining "remedy" as "[t]he means by which a right is enforced or the violation of a right is prevented, redressed, or compensated." Id . at 1086 (quoting Remedy , Black's Law Dictionary 1526 (3d ed. 1933)) (quotation marks omitted). And it asserts that "remedy" does not necessarily equate with "relief." See id . I have no quarrel with this definition up to this point.

In fact, I agree with it. And the interpretation of the saving clause that I offer is entirely consistent with this definition. Under the language of § 2255, the "right [to be] enforced," see id. is the right protected by the Suspension Clause not to be detained in violation of the separation of powers or the principle of limited government powers, through the impermissible interpretation of the statute or other device used to detain the prisoner, that has since been invalidated by the Supreme Court's retroactively applicable new rule of substantive law—whether constitutional or statutory. See supra at 1132–40.

Regardless of whether the Supreme Court's new substantive rule is characterized as constitutional or statutory in nature, the right under which access to that rule is claimed is constitutional—a constitutional right protected by the Suspension Clause not to be detained in excess of the government's powers. And, as a practical matter, that constitutional right, which is based on a new rule of substantive law made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable, cannot be vindicated until after the Supreme Court announces the new rule.²⁷

27 The Majority asserts that "this argument ignores that litigants often make novel arguments in the hope that a court will adopt them as a matter of first impression or in a rejection of past precedent" and that "[i]t is unclear why the chance to have precedent overruled en banc or by the Supreme Court would not qualify as a theoretically successful challenge or meaningful opportunity." Maj. Op. at 1087, 1091. This argument misses the point. The fact that litigants may raise novel arguments to courts before such time that the Supreme Court has issued a new rule of substantive law made retroactive to cases on collateral review, that was previously unavailable, does not mean that litigants have an opportunity at all—let alone a meaningful opportunity—to invoke a right that is based on a new retroactively applicable rule of substantive law issued by the Supreme Court, which was not recognized prior to the Supreme Court's announcement of it. How could litigants when the Supreme Court had not yet recognized the very rule on which their right relies?

But § 2255, in the absence of the saving clause, does not provide a "remedy," or "means by which a right is enforced," Maj. Op. at 1086, when it comes to new rules of statutory law. For this reason, § 2255 (without the saving clause) fails to provide a "meaningful opportunity" for a prisoner to test the legality of his detention when the Supreme Court issues a retroactively applicable new rule of statutory law. See supra at 1135–40. As a result, the construction of the saving clause that I put forward comports with the Majority's initial proposed definition of "remedy."

Next, the Majority initially defines "to test" as meaning "to try." Maj. Op. at 1086–87 (citing Test , 11 Oxford English Dictionary 220 (1st ed. 1933)). It then turns to Prost to argue that in order "to test," a "petitioner [must have] an opportunity to bring his argument." Id. (quoting Prost , 636 F.3d at 584 ) (quotation marks omitted). Setting aside for the moment the facts that Prost expressly declined to consider constitutional issues in its interpretation, see supra at 1152, and that "an opportunity" must be a "meaningful opportunity," see id. at 1135–40, I do not take issue with this construction.

And as with the Majority's dictionary definition of "remedy," the construction of the saving clause that I offer in this dissent is consistent with the Majority's initial definition of "to test." A prisoner whose argument depends on a "right [that] has been newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review" has no argument based on that right until the Supreme Court has newly recognized the rule on which the right invoked is based and rendered that rule retroactively applicable. So that prisoner has no opportunity—and certainly no meaningful opportunity—to test the legality of his detention until after the Supreme Court announces a new retroactively applicable rule of substantive law, whether constitutional or statutory.

Nor do I object to the Majority's initial definition of "detention," which it defines as "[k]eeping in custody or confinement," Maj. Op. at 1089 (quoting Detention , 3 Oxford English Dictionary 266 (1st ed. 1933)), and "[t]he act of keeping back or withholding, either accidentally or by design, a person or thing," id. (quoting Detention, Black's Law Dictionary 569 (3d ed. 1933)). Once again, the interpretation of the saving clause presented in this dissent comports with these definitions of "detention" and ascribes a different meaning to "sentence" and "detention" as used in § 2255. See supra at 1129–30 (citing Brown, 817 F.3d at 1284 ). As I have explained, testing the "legality of detention" means bringing a claim that, if correct, would result in a reduced period of detention, not just a reduced sentence on a given count that may or may not affect the overall period of detention.

But the Majority's explanation of these terms' meanings flies off the rails when the Majority contorts its original version of its definitions to fit its vision of the saving clause's meaning. For instance, in interpreting "remedy" and "to test" to mean that the saving clause precludes a claim when a prisoner had an opportunity under § 2255 to raise an argument on an issue—as opposed to a right based on a new retroactively applicable rule, see Maj. Op. at 1086–87—the Majority ignores the meaning of "right" as that term is used in § 2255(f)(3), see supra at 1150–51, and proves too much.

Under the Majority's construction, in the absence of subsection (h)(2), the saving clause would not allow a prisoner to bring a second or successive claim based on a new retroactively applicable rule of constitutional law because the prisoner would have had the opportunity to have made the argument on the same general issue on his initial § 2255 motion, before the Supreme Court announced a new retroactively applicable rule of constitutional law on that issue. Necessarily, then, in the Majority's view, Congress could have chosen to ban second or successive claims based on a retroactively applicable new rule of constitutional law.

But Congress's inclusion of the subsection (h)(2) exception to the ban on second or successive motions was not optional. The Suspension Clause required it. See supra at 1133–40. As we have discussed, the separation-of-powers doctrine and the principle of limited government powers "must inform the reach and purpose of the Suspension Clause." Boumediene , 553 U.S. at 746, 128 S.Ct. 2229 (emphasis added). And "the 'retroactivity' of a new constitutional rule [is] a function of the scope and purposes of the habeas corpus writ," Mackey , 401 U.S. at 684, 91 S.Ct. 1160 (Harlan, J., concurring), so new constitutional rules are retroactively applicable on collateral review because the Suspension Clause requires it, not because Congress was in a generous mood when it enacted AEDPA. Any construction of the saving clause that does not acknowledge this fact does not account for the saving clause's function to save § 2255 from unconstitutionality.

As for the Majority's reinterpretation of (as opposed to initial citation to) the dictionary definitions for "detention," see Maj. Op. at 1089–90, it relies on no sources that support it,²⁸ other than Judge Pryor's own concurrence in Samak , and it conflicts directly with the plain language of § 2255. See supra at 1147–49. For this reason, it cannot be correct.

28 The Majority cites several cases, but they do not stand for the proposition that execution-of-sentence claims are cognizable under the saving clause. See Maj. Op. at 1089–90. On the contrary, without mentioning the saving clause in any way, they assert that execution-of-sentence claims are not appropriately brought under § 2255 but rather under § 2241—my point exactly.

Finally, and perhaps most glaring of all, the Majority's definition of "inadequate or ineffective" entirely fails to account for the saving clause's constitutional-failsafe function. We know that "[t]he [Supreme] Court placed explicit reliance upon [the saving clause] provisions in upholding [28 U.S.C. § 2255 and the District of Columbia equivalent of § 2255 ] against constitutional challenges." Boumediene , 553 U.S. at 776, 128 S.Ct. 2229 (citing Swain , 430 U.S. at 381, 97 S.Ct. 1224 ; Hayman , 342 U.S. at 223, 72 S.Ct. 263 ). If, as the Majority's construction suggests, the term "ineffective" does not require consideration of claims that the Suspension Clause requires be heard, what part of the saving clause authorizes consideration of constitutional challenges when no other part of § 2255 does? The Majority offers no alternative, even though the Supreme Court has repeatedly acknowledged § 2255's constitutional-failsafe function. See id .

For all of these reasons, the Majority's interpretation of the saving clause cannot be correct.

IV.

Chief Judge Carnes's concurrence asserts that I engage in judicial activism²⁹ in this dissent, " 'improv[ing]' the statute by writing in the exception that [I] favor[ ]." E. Carnes Op. at 1100. As always with the Chief, his concurrence is beautifully written. So it would be easy to succumb to its seductive Siren song without considering whether the comments in the concurrence are, in fact, correct.

29 The Chief's concurrence takes issue with my characterization of its criticism as charging that I have engaged in judicial activism. See E. Carnes Op. at 1100–01. So I pause to explain why I describe its criticism that way. True, the concurrence never actually employs the phrase "judicial activism." Instead, it describes my dissent as having " 'improve[d]' the statute by writing in the exception that [I] favor [ ]," "amend[ed] the statute," engaged in an activity that is not in line with "the proper role of the judiciary," "judicial[l]y revis[ed] ... [the] statute[ ]," imposed my "musings, whether pragmatic or otherwise," on the statute, "design[ed]" a statute, and "rewrit[ten]" the statute. Id. at 1100–01. My goodness! The concurrence sure thinks I've been very busy doing our legislators' jobs. And "[l]egislating from the bench ... [is just] another name for judicial activism." Thomas L. Jipping, Legislating From the Bench: The Greatest Threat to Judicial Independence, 43 S. Tex. L. Rev. 141, 146 (2001).

But let's take a moment to think about the concurrence's premise. In the concurrence's view, I have "added a third exception" to subsection (h)'s bar on second or successive claims, and I have done so because that is the result I allegedly desire. Id. at 1100. So as the concurrence sees things, before we even start our analysis of the saving clause, we necessarily must agree to a ground rule that the saving clause cannot have a constitutional-failsafe purpose as it relates to second or successive claims. For if it does and any constitutionally required second or successive claims are not accounted for by subsection (h)'s exceptions to the bar on second or successive claims, the judge who observes this deficiency must be a judicial activist.

In an ironic twist, though, the concurrence's approach itself embodies judicial activism: instead of reviewing the text and relevant precedents and seeing where they take us, it begins with an end in mind before analysis even starts—and it does so despite the fact that the Supreme Court has reminded us on more than one occasion that the saving clause must have a constitutional-failsafe function to protect § 2255 from unconstitutionality. See Boumediene , 553 U.S. at 776, 128 S.Ct. 2229 (citing Swain , 430 U.S. at 381, 97 S.Ct. 1224 ; Hayman , 342 U.S. at 223, 72 S.Ct. 263 ).

Even if we were to accept the concurrence's accusation at face value, exactly when does the analysis in this dissent engage in judicial activism? When it spends seventeen pages parsing the text, grammar, and function of § 2255 ? When it relies on the Supreme Court cases Hayman , Swain , and Boumediene for the proposition that the saving clause acts as a constitutional failsafe to protect § 2255 from unconstitutionality? When it invokes Boumediene , The Federalist No. 84, and Hamdi to show that the separation-of-powers doctrine and the principle of limited government powers drive habeas and the Suspension Clause? Or perhaps when this dissent cites Justice Harlan's concurrence in Mackey and the Supreme Court's opinions in Bousley and Welch to show that retroactivity doctrine—including retroactivity doctrine as it relate to new rules of statutory law—is required by the same separation-of-powers and limited-government concerns that animate habeas and the Suspension Clause?

Maybe the judicial activism occurs when this dissent suggests the Majority's contention that execution-of-sentence and pretrial detention claims may be considered under the saving clause conflicts with the language and function of § 2255. I don't know. And the reason I don't know is that the Chief's concurrence does not direct its criticism to any particular step in my analysis, instead just asserting that I have added a third exception to § 2255's bar on second or successive motions.

I acknowledge that this dissent may present a new theory on why the saving clause requires consideration of second or successive claims based on a retroactively applicable new rule of statutory law—though, of course, it is not the first opinion to conclude that second or successive claims based on a retroactively applicable new rule of statutory law may be considered under the saving clause. See, e.g. , Wofford , 177 F.3d at 1244 (E. Carnes, J.);³⁰ Triestman v. United States , 124 F.3d 361, 363 (2d Cir. 1997) ; In re Dorsainvil , 119 F.3d 245, 251 (3d Cir. 1997) ; In re Jones , 226 F.3d 328, 333-34 (4th Cir. 2000) ; Reyes-Requena v. United States , 243 F.3d 893, 904 (5th Cir. 2001) ; Wooten v. Cauley , 677 F.3d 303, 307-08 (6th Cir. 2012) ; In re Davenport , 147 F.3d 605, 611-12 (7th Cir. 1998) ; In re Smith , 285 F.3d 6, 8 (D.C. Cir. 2002).

30 The Chief's concurrence falls on its Wofford sword in the most eloquent fashion. See E. Carnes Op. at 1101. Though I can't help but admire and be entertained by the Chief's way with words, I see no reason for the concurrence to apologize for Wofford or its progeny. I do not aim to impose blame on the Chief or this Court for our prior jurisprudence on the saving clause. Our understanding of the law should develop in response to Supreme Court precedent, and I agree with the concurrence that we should reconsider our views when another demonstrates the incorrectness of an existing precedent. I also share the concurrence's position that our analysis of the saving clause should not start from Wofford 's analysis and "revis[e]" it. Id. at 1101. Instead, our analysis must begin with the statutory text and function, as informed, where the text is ambiguous, by Supreme Court precedent. And it should end wherever that analysis may take us—without some preexisting idea of where that should be. I cite Wofford for only the point that my ultimate conclusion—that second or successive claims based on a retroactively applicable new rule of statutory law may be considered under the saving clause—is not novel.

Presenting a new theory, however, is not the same thing as engaging in judicial activism. If it were, we could not have circuit splits unless judicial activism occurred—any circuit that arrived at an interpretation contrary to the first court's resolution of the issue would necessarily be engaging in judicial activism just by proposing a new theory. But the mere fact that judges may disagree over the meaning of a provision and may offer an alternative theory to explain the basis for the disagreement does not mean that a judge is engaged in judicial activism.

Instead, we evaluate that by how the new theory is supported. We must consider whether the new theory is based on a fair interpretation of the statutory text and binding precedent. I respectfully submit that this dissent's theory is.

This dissent relies on the plain meaning of § 2255's text, to the extent that it is unambiguous. And to the extent that it is not, this dissent reviews, considers, and then simply points out existing, though perhaps not previously observed, lines among Supreme Court cases to inform the meaning of the saving clause's constitutional-failsafe function. That is not judicial activism; it is legal analysis.

V.

The range of interpretations courts—including ours—have applied to § 2255(e) may make construing the saving clause seem like the Kobayashi Maru³¹ of law training exercises. But, actually, the saving clause has a single correct meaning, and the secret to understanding it lies in the text of the clause, as informed by the constitutional-failsafe function of the clause. That meaning requires consideration of second or successive claims, the failure of which to consider would render § 2255 constitutionally deficient. Second or successive claims based on retroactively applicable new rules of statutory law announced by the Supreme Court fall within this category. Since McCarthan invokes such a rule, I would reverse the denial of his claim and remand for the district court to consider the merits in the first instance. Because we don't do that, I respectfully dissent.

31 In the Star Trek universe, Kobayashi Maru is a training exercise for Starfleet Academy cadets. In it, the cadet must determine whether to attempt a rescue of the Kobayashi Maru, a disabled Starfleet ship, risking death to the rescuers, or whether instead to decline the rescue of the Kobayashi Maru, risking death to those onboard the stranded vessel. Though cadets taking the training exercise are under the impression that it tests their strategic skills as ship commanders, unbeknownst to the cadets, the exercise is designed as an unwinnable scenario and is administered for the purpose of testing the character of the cadets. See Star Trek (Paramount Pictures, Spyglass Entertainment, Bad Robot, Mavrocine Pictures GmbH & Co. KG 2009); Star Trek II: The Wrath of Khan (Paramount Pictures 1982).
--------