This case is one of three companion eases that we decide today; the three cases address interrelated issues bearing on prisoner litigation.
In this action, Clarence Blair-Bey, a prison inmate in the custody of the D.C. Department of Corrections, filed a habeas corpus petition challenging the procedures by which he was denied parole, on due process and ex post facto clause grounds. He named as defendants (among others) the D.C. Board of Parole and the United States Parole Commission. The district court dismissed Blair-Bey’s petition, finding that the federal courts are precluded by a provision of the D.C.Code, section 16-1901, from entertaining habeas corpus petitions filed by D.C. prisoners, and that Blair-Bey’s claims also failed on their merits. Blair-Bey appealed these two rulings. In this court, the District defends those rulings, and also contends that Blair-Bey is required, under the filing-fee provisions of the recently-enacted Prison Litigation Reform Act (“PLRA”), 28 U.S.C. § 1915(d), to pay a partial filing fee before his appeal may proceed. The United States Parole Commission agrees with the District on the grounds for dismissal but not as to the applicability of the PLRA’s filing fee provisions; it also asserts that it should be dismissed as a party.
We conclude that Blair-Bey’s action, as a habeas corpus petition, is not subject to the PLRA’s filing-fee requirements; that this court has jurisdiction to entertain Blair-Bey’s petition; that the United States Parole Commission is not a proper defendant; and that although Blair-Bey’s due process claim is meritless, his ex post facto claim might have merit if certain facts are shown, and must be remanded to the district court for further consideration.
I. Background
Blair-Bey was born in 1958. Between the ages of thirteen and fifteen, he committed a number of serious offenses, including rape, receiving stolen property, destruction of property, and two counts of unauthorized use of a motor vehicle. As a result, he spent much of his youth at the District of Columbia’s juvenile detention facility, Cedar Knoll. Blair-Bey also had psychological and substance-abuse problems during this period, and so was confined part of this time at St. Elizabeths Hospital. In 1975, at the age of sixteen, he escaped from Cedar Knoll and committed a murder. Appendix (“App.”) at 31, 44. Blair-Bey pleaded guilty to a D.C.Code offense of second-degree murder, and was given a sentence of from ten years to life. In 1980, while serving his sentence, he killed another inmate; he pleaded guilty to second-degree murder in violation of 18 U.S.C. § 1111, and was sentenced to a term of ten years to life to be served consecutively with his prior sentence. He was then transferred to a federal correctional facility and *1038began to serve his second (U.S.Code) sentence.
In 1991, the United States Parole Commission (“USPC”) held a parole hearing on Blair-Bey’s federal offense, denied parole, and scheduled another hearing for 1993. The USPC said in its order that at the next proceeding Blair-Bey would also receive a parole hearing for his D.C.Code offense. The USPC and the D.C. Board of Parole (“DCBOP”) have an arrangement under which the USPC conducts parole hearings (under D.C. parole rules) for those D.C.Code offenders being held in federal prison. ■
Blair-Bey finished serving the federal portion of his sentence on October 5, 1991, and was transferred back-to a D.C. prison in 1993. Apparently in error, a USPC hearing examiner nevertheless conducted a parole hearing for him at the D.C. Jail on September 8, 1993. The examiner found that Blair-Bey had a favorable score under the D.C. scoring system (a “point assigned grid score,” or PAGS, of 2, which corresponds to “parole shall be granted”), but nevertheless recommended to the USPC that Blair-Bey be denied parole. The hearing examiner cited the seriousness of Blair-Bey’s crimes, his lack of remorse, his drug use while in prison, and lies he told about his movements while on work-release. App. at 34-36. The examiner recommended that Blair-Bey receive a rehearing in September 1995.
Before the USPC could formally adopt these recommendations, the USPC and DCBOP decided that it was in fact the DCBOP that had jurisdiction over Blair-Bey. The USPC sent a copy of its file and recommendations to the DCBOP, which conducted a de novo parole-hearing on October 29,1993. Unlike the USPC, the DCBOP found that Blair-Bey had an unfavorable score under the D.C. system, giving him a PAGS score of 4, which corresponds to denying parole. (The DCBOP seems to have counted Blair-Bey’s numerous juvenile offenses in making its calculation, App. at 47-48; the USPC did not count these offenses.) Citing his use of drugs, work-release misconduct, and diagnosis (while in prison) of.paranoid schizophrenia, the hearing examiner decided that Blair-Bey was a threat to the community. He recommended that parole be denied, and that no new hearing be scheduled until October 29, 1998. This represented a five-year “set-off,” or delay until Blair-Bey’s next hearing, longer than the usual set-off of twelve months (or less) provided for in the D.C. guidelines. See D.C. Mun. Regs. tit. 28, § 104.2, 35 D.C.Reg. 455 (1988). The DCBOP adopted this recommendation in full, noting (through checkmarks on a form) that Blair-Bey’s convictions reflect “on-going criminal behavior” and that Blair-Bey “has engaged in repeated or extremely serious negative institutional behavior.” The DCBOP’s form also noted that “[t]he set-off is outside the guidelines for the same countervailing factors checked below.” App. at 56..
Blair-Bey filed a habeas corpus petition in D.C: Superior Court, claiming that the DCBOP acted illegally in denying him parole and in establishing a five-year set-off. The Superior Court denied the petition, and the D.C. Court of Appeals summarily affirmed. Blair-Bey v. Quick, No. 96-1593, slip op. at 1 (D.C. Sept. 28, 1995). Blair-Bey then filed a similar petition in federal district court, fist-ing as defendants the warden of Occoquan, the chair of the DCBOP, his DCBOP hearing examiner, and the USPC.
The district court dismissed Blair-Bey’s petition sua sponte for lack of jurisdiction, concluding that, under D.C.Code § 16-1901, only the courts of the District of Columbia have jurisdiction over petitions like Blair-Bey’s. In the alternative, the district court found that Blair-Bey’s claims failed on their merits, as he had no constitutional right to release on parole.
Blair-Bey appealed, challenging both of these conclusions. Blair-Bey’s appeal was argued together with two other appeals brought by prison- inmates, Anyanwutaku v. Moore, No. 96-7259, and Crowell v. Walsh, No. 96-7192, because they raised related issues; we decide all three of these cases (in separate opinions) today. The District asserts, first, that Blair-Bey’s appeal is subject to the filing-fee requirements of the PLRA, and that Blair-Bey should therefore be required to pay a (partial) fifing fee before we proceed. We reject this claim, finding that *1039Biair-Bey’s action, as it is a habeas petition, is outside the filing-fee provisions of the PLRA. ' Second, the District contends that the district court was correct in concluding that the federal courts lack jurisdiction over Biair-Bey’s petition; again, Blair-Bey wins the day. Finally, as to the merits of Blair-Bey’s claims, we find that his due process claim is meritless, but that his ex post facto claim requires further development on remand.
II. Analysis
At the outset, it is clear that the United States Parole Commission is not a proper party to this action. “[T]he appropriate defendant in a habeas action is the custodian of the prisoner.” Chatman-Bey v. Thornburgh, 864 F.2d 804, 810 (D.C.Cir.1988) (en banc). When a prisoner seeks to challenge parole-related decisions, the warden of the prison and not the United States Parole Commission is the prisoner’s “custodian.” See id. at 810-11 (citing Guerra v. Meese, 786 F.2d 414, 416 (D.C.Cir.1986)). Thus, the USPC, the DCBOP, and their officers and employees are not proper parties to this action; only John S. Henderson, the warden of Occoquan, is a proper party.1
A. The Prison Litigation Reform Act
In 1996, Congress enacted the Prison Litigation Reform Act, Pub.L. No. 104-134 (1996), which made a number of changes in the law governing prison-related litigation. The portion of the PLRA that concerns us here is section 804, which amended the statute governing proceedings in forma pauperis (“IFP”) to require that, “if a prisoner brings a civil action or files an appeal in forma pauperis,” he must pay the fee over time; payments are deducted from the plaintiffs prison account. 28 U.S.C. § 1915(b)(l)-(3). (A prisoner -will not, however, be prevented from filing suit simply because he has no assets. 28 U.S.C. § 1915(b)(4).)
We have found, following numerous other circuits, that Congress did not intend for this installment-payment provision to apply to actions brought under 28 U.S.C. § 2254 or § 2255. See United States v. Levi, 111 F.3d 955, 956 (D.C.Cir.1997) (per curiam) (citing eight opinions of other circuits); In re Smith, 114 F.3d 1247, 1250 (D.C.Cir.1997). As we explained in In re Smith, the courts have “uniformly concluded” that habeas corpus proceedings — and their cousins, section 2255 proceedings — are not “civil actions” for purposes of the PLRA’s filing-fee provision. See id.; see also Smith v. Angelone, 111 F.3d 1126, 1129-30 (4th Cir.), cert. denied, - U.S. -, 118 S.Ct. 2, 138 L.Ed.2d 1036 (1997) (§ 2254); United States v. Simmonds, 111 F.3d 737, 742 (10th Cir.1997) (§ 2255).
The District argues, however, that we should apply Levi only to those habeas corpus petitions that challenge the validity of the petitioner’s conviction or sentence, and not to petitions, like this one, that are based on other grounds. As authority, it cites Newlin v. Helman, 123 F.3d 429 (7th Cir.1997), cert. denied, - U.S. -, 118 S.Ct. 707, 139 L.Ed.2d 649 (1998), in which the Seventh Circuit distinguished between (1) petitions attacking the prisoner’s conviction or sentence, which it said were not “civil actions” for PLRA purposes because they are “functional continuations of the criminal prosecution,” and (2) “[c]omplaints about denial of parole, revocation of parole, and the like,” which “do not affect the validity of the criminal sentence” and hence may be treated as “civil actions.” Id. at 437-38. Blair-Bey and the United States Parole Commission urge us not to follow Newlin.
At least one other circuit has already rejected Newlin (or rather a pre-Newlin case, Thurman v. Gramley, 97 F.3d 185, 187 (7th Cir.1996), which laid out Newlin’s approach in dicta); See McIntosh v. United States Parole Commission, 115 F.3d 809, 811-12 (10th Cir.1997) (finding that a habeas corpus challenge to a parole revocation proceeding was not subject to the PLRA). We agree *1040with McIntosh-, we see no reason to make an exception from our general rule that the PLRA does not apply to actions properly brought in habeas corpus.
By its terms, the PLRA only applies when “a prisoner brings a civil action or files an appeal in forma- pauperis.” 28 U.S.C. § 1915(b)(1).2 The courts have generally found habeas corpus petitions not to be “civil actions” for purposes of this provision. “Though habeas proceedings are technically civil actions, the Supreme Court has long recognized that the label is ill-fitting and that habeas is in fact a unique creature of the law.” Martin v. Bissonette, 118 F.3d 871, 874 (1st Cir.1997) (citing Harris v. Nelson, 394 U.S. 286, 293-94, 89 S.Ct. 1082, 22 L.Ed.2d 281 (1969)). As the Court explained in Hams,the characterization of habeas proceedings as “civil” is “gross and inexact.” 394 U.S. at 293-94, 89 S.Ct. 1082. “Essentially, the proceeding is unique. Habeas corpus practice in the federal courts has conformed with civil practice only in a general sense.” Id. at 294, 89 S.Ct. 1082. More recently, in O’Neal v. McAninch, 513 U.S. 432, 115 S.Ct. 992, 130 L.Ed.2d 947 (1995), the Court found that it is error to apply standards of proof developed for civil actions to a habeas corpus claim, because of “the stakes involved in a habeas proceeding”; “although habeas is a civil proceeding, someone’s custody, rather than mere civil liability, is at stake.” Id. at 440, 115 S.Ct. 992; see also Santana v. United States, 98 F.3d 752, 754-55 (3d Cir.1996) (listing numerous circumstances in which habeas corpus proceedings are not treated as civil actions). None of these cases draw any distinction between different types of habeas petitions that resembles the “eriminar/“civil” distinction made by Newlin, and Newlin cited no authority for this distinction.
There is thus no evidence that Congress might have relied on a preexisting distinction between “criminal” and “civil” habeas corpus petitions when it enacted the PLRA. Nor is there any indication that Congress itself intended to establish any such distinction in the PLRA. The PLRA’s legislative history makes clear that Congress’s principal intent was to reduce frivolous litigation by prisoners challenging conditions of their confinement. Although there was no committee report on the PLRA, the floor statements of the bill’s two sponsors in the Senate, Senators Dole and Kyi, indicate that prison conditions litigation is what they were targeting. All examples of suits they hoped to squelch were prison conditions suits, including lawsuits challenging “insufficient storage space, being prohibited from attending a wedding anniversary party, and, yes, being served creamy peanut butter instead of chunky.” 141 Cong. Rec. S7498, S7524 (daily ed. May 25, 1995) (statement of Senator Dole); see also In re Smith, 114 F.3d at 1249 ('“Congress enacted the PLRA primarily to curtail claims brought by prisoners under 42 U.S.C. § 1983 and the Federal Tort Claims Act, most of which concern prison conditions and many of which are routinely dismissed as legally frivolous.”) (quoting Santana, 98’F.3d at 755). Nowhere in the PLRA’s legislative history is there any reference to cases challenging the fact or duration of confinement.3 Indeed the PLRA contains several other provisions directed specifically at prison conditions litigation, see, e.g., PLRA § 802 (relating to prison conditions consent decrees); § 803(d) (requiring exhaustion in actions “with respect to prison conditions”), but makes no specific reference anywhere in the Act to challenges to the fact or length of confinement.
*1041Just two days before enacting the PLRA, Congress enacted a distinct statute, the Anti-terrorism and Effective Death Penalty Act (“AEDPA”), Pub.L. No. 104-132 (1996), which made extensive changes in the law governing habeas relief. “This chronology strongly suggests that Congress intended to make its changes to habeas proceedings via the AEDPA, and to alter procedure in prisoner civil rights litigation in the PLRA” Smith v. Angelone, 111 F.3d at 1130; see also In re Smith, 114 F.3d at 1250. Although petitions attacking criminal convictions were the principal focus of the AEDPA, a number of the AEDPA’s provisions apply on their face to all habeas petitions, and so would seem to reach petitions like Blair-Bey’s.
Treating one subset of habeas petitions as “civil actions” for PLRA purposes would also have the effect of subjecting those petitions to two separate regimes designed to deter repeat plaintiffs — with anomalous results, given the nature of the two regimes. Even before the AEDPA and PLRA were enacted, 28 U.S.C. § 2244 limited “second or successive” habeas corpus petitions; section 106 of the AEDPA adjusted these limitations, and so recognized the existence of a repeat-filer regime directed specifically at habeas corpus petitions. The PLRA has introduced a new repeat-filer provision, providing that a prisoner who has had three suits dismissed on the ground that they were “frivolous, malicious, or fail[ed] to state a claim upon which relief can be granted” will be required to prepay filing fees in full in future civil actions or appeals, unless the prisoner can show “imminent danger of serious physical injury.” 28 U.S.C. § 1915(g). This means that there are now two distinct regimes for repeat filers, and it seems to us that the PLRA’s regime is singularly ill-suited to habeas corpus petitioners. If habeas petitions count as “civil actions” for PLRA purposes, then a penniless inmate 4 who has managed to incur three strikes would apparently have no way of seeking a writ of habeas corpus — even if, for instance, prison officials arbitrarily declined to release him at the end of his sentence. This would truly be a draconian penalty, and “contrary to a long tradition of ready access of prisoners to federal habeas corpus, as distinct from their access to tort remedies.” Anderson v. Singletary, 111 F.3d 801, 805 (11th Cir.1997) (quoting Martin v. United States, 96 F.3d 853, 855-56 (7th Cir.1996)); see also O’Neal, 513 U.S. at 432, 440, 115 S.Ct. 992 (1995) (observing that the stakes in habeas proceedings are high, as custody over an individual is at issue). We will not lightly assume that Congress intended such an extraordinary result.5
The District asserts that it can often be difficult .to determine whether. a prisoner’s action is properly in habeas corpus, and argues that an approach that does not limit the PLRA’s exemption to hard-core habeas actions will compel courts to make this decision very early, before necessary facts have been developed. It is true that this line can be a hazy one, and the Supreme Court has had to revisit it a number of times in recent years. See,. e.g., Edwards v. Balisok, 520 U.S. 641, 117 S.Ct. 1584, 137 L.Ed.2d 906 (1997). But Newlin’s line between “criminal” and “civil” habeas petitions seems almost as elusive. (How should a habeas corpus petition by a prisoner who is being held for contempt be classified? Or one by a pretrial detainee?) Nor do we think our reading imposes an inordinate taxonomic burden on the district courts. The issue of whether an action sounds in habeas corpus will often come up early in the case anyway, for if an action by a state prisoner is truly in habeas corpus, the petition is subject to threshold exhaustion requirements, see 28 U.S.C. § 2254. And many cases, such as pure prison-conditions cases, will be easy to identify. Furthermore, if a prisoner joins a claim that clearly cannot be heard in habeas corpus (such as a claim for damages) along with a claim that argu*1042ably may be, the resulting “mixed” petition will always be subject to the PLRA’s filing-fee rules. See In re Smith, 114 F.3d at 1250. For the rare suit that raises truly bewildering issues of classification, it may be appropriate for the district courts to conserve their resources by deferring the .(potential) application of the PLRA’s installment-payment provisions until the facts of the case can be better developed.
There remains one other classification issue.. It is possible that habeas corpus might be-available to challenge prison conditions in at least some situations. The Court expressly left this possibility open in Preiser v. Rodriguez, see 411 U.S. 475, 499, 93 S.Ct. 1827, 36 L.Ed.2d 439 (1973); see also Brown v. Plant, 131 F.3d 163, 168 (D.C.Cir.1997), cert. denied, - U.S. -, 118 S.Ct. 2346, 141 L.Ed.2d 716 (1998); Abdul-Hakeem v. Koehler, 910 F.2d 66, 69-70 (2d Cir.1990); but cf. Gomez v. United States, 899 F.2d 1124, 1125-26 (11th Cir.1990). Such claims, if they are permissibly brought in habeas corpus, would have to be subject to the PLRA’s filing fee rules, as they are precisely the sort of actions that the PLRA sought to address. See In re Smith, 114 F.3d at 1250 (D.C.Cir.1997) (“[I]t would defeat the purpose of the PLRA if a prisoner could evade its requirements simply by dressing up an ordinary civil action as a petition for mandamus or prohibition or by joining it with a petition for habeas corpus.”)
We thus reject the District’s suggestion that we apply the PLRA’s filing-fee provisions to habeas petitions like the one in this case, and now turn to the District’s D.C.Code argument.
B. D.C.Code § 16-1901
The District asserts that we are without jurisdiction to hear Blair-Bey’s habeas petition, because such petitions may only be brought in the courts of the District of Columbia. It bases this argument on a provision of the D.C.Code, section 16-1901, which, it claims, divests the federal courts of jurisdiction to hear habeas corpus petitions filed by D.C. prisoners.
1. The problem
Some historical and statutory background is necessary to parse the District’s argument. Before 1970, the D.C. court system did not exist in its present form, and many of the cases now brought in the District’s courts were instead heard in federal court. In 1970, Congress passed the District of Columbia Court Reform and Criminal Procedure Act, Pub.L. No.' 91-358 (1970) (“DCCRCPA”), which created the present dual court system.
.In the course of creating this new dual system, the DCCRCPA established a remedy analogous to 28 U.S.C. § 2255 for prisoners sentenced in D.C. Superior Court who wished to challenge their conviction or sentence. See D.C.Code § 23-110. The DCCRCPA provided that, to the extent that this remedy was available, it was an exclusive one:
An application for a writ of habeas corpus on behalf of a prisoner who is authorized to apply for relief by motion pursuant to this section shall not be entertained by the Superior Court or by any Federal or State court if it appears that the applicant has failed to make a motion for relief under this section [§ 23-110] or that the Superior 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.
D.C.Code § 23-110(g). Thus, the DCCRCPA entirely divested the federal courts of jurisdiction to hear habeas corpus petitions by prisoners who had a section 23-110 remedy available to them, unless the petitioner could show that the section 23-110 remedy was “inadequate or ineffective,” an exception that we will call the “safety valve.” In Swain v. Pressley, 430 U.S. 372, 97 S.Ct. 1224, 51 L.Ed.2d 411 (1977), the Supreme Court upheld the constitutionality of section 23-110(g) under the Suspension Clause, opining that the safety valve “avoids any serious question about the constitutionality of the statute.” See id. at 381, 97 S.Ct. 1224. The Court explained that “the substitution of a collateral remedy which is neither inadequate nor ineffective to test the legality of a person’s detention does not constitute a suspension of the writ of habeas corpus.” See id. *1043(observing that the same is true of 28 U.S.C. § 2255, which contains an identical safety-valve).
Blair-Bey’s petition seeks his release, but does not challenge his conviction or sentence; thus, it could not be brought under section 23-110. See Alston v. United States, 590 A.2d 511, 514 (D.C.1991) (“Whatever their legal merit, these contentions, like claims by other prisoners challenging the computation of a sentence, may not be raised under § 23-110. Because such contentions concern the executive department’s execution of sentence, not the trial court’s imposition of sentence, they must be raised in a habeas corpus petition.”). Because section 23-110(g) only bars us from hearing those claims that could have been raised through section 23-110, that section does not prevent us from entertaining Blair-Bey’s habeas petition. The District, however, points to a separate provision of the D.C.Code, which it claims does preclude us from considering Blair-Bey’s petition. That provision, D.C.Code § 16-1901, states:
(a) A person committed, detained, confined or restrained from his lawful liberty within the District, under any color or pretense whatever, or a person in his behalf, may apply by petition to the appropriate court, or a judge thereof, for a writ of habeas corpus, to the end that the cause of the commitment, detainer, confinement, or restraint may be inquired into. The court or judge applied to, if the facts set forth in the petition make a prima facie case, shall forthwith grant the writ, directed to the officer or other person in whose custody or keeping the party so detained is, returnable forthwith before the court or judge.
(b) Petitions for writs directed to Federal officers and employees shall be filed in the United States District Court for the District of Columbia.
(c) Petitions for writs directed to any other person shall be filed in the Superior Court for the District of Columbia.
In its present form, this provision dates from the enactment of the DCCRCPA. It sets forth a general habeas corpus remedy for those confined in the District of Columbia, allowing prisoners to attack their confinement on any grounds (other than those provided for in section 23-110).6 The District contends that, because Blair-Bey is confined in the District and because his petition is not directed to a federal officer or employee, paragraph (c) of section 16-1901 directs that he file his habeas corpus petition in D.C. Superior Court, and bars him from filing it in federal district court. The district court agreed with that argument and other cases in this circuit have come to the same conclusion. See, e.g., Perkins v. Henderson, 881 F.Supp. 55 (D.D.C.1995).
A preliminary point: At the time the district court issued its order, Blair-Bey was imprisoned at the District’s Occoquan facility, which is not within the territorial limits of the District. But the term “within the District” in section 16-1901 is one of art. In McCall v. Swain, 510 F.2d 167 (D.C.Cir. 1975), we read that phrase to encompass “individuals confined within the District’s correctional facilities located outside the District limits.” Id. at 177. For purposes of section 16-1901, then, Blair-Bey was “within the District” at the time the district court issued its order.
It may seem curious that the question of whether the federal courts can entertain a habeas corpus petition by a D.C. prisoner who does not seek to challenge his conviction or sentence should have remained unanswered in the decades since the DCCRCPA was enacted. But it has, and today we answer it. We find that section 16-1901 does not bar the federal courts from entertaining habeas corpus petitions filed by D.C. prisoners under 28 U.S.C. § 2241. Sections 16-1901(b) and (c) only set forth the proper place in which to file those habeas corpus petitions that are brought pursuant to section 16-1901(a). They do not speak to the question of where persons may file habeas petitions that are brought under other sources of authority, such as section 2241.
In deciding whether Congress intended to restrict the availability of federal habeas cor*1044pus when it enacted the present section 16-1901, we tread carefully. As reviewed above, there is a “long tradition of ready access of prisoners to federal habeas corpus,” Anderson, 111 F.3d at 805, and we are most reluctant to find that Congress has .deprived an entire category of prisoners of access to an Article III habeas remedy without very clear evidence of congressional intent. As section 23-110(g) demonstrates, Congress certainly knew how to express itself .clearly on the exclusivity of a remedy when it wanted to do so.
2. The statutory text
We begin with the statutory text. In contrast to section 23-110(g), section 16-1901 does not express in any form an intent to eliminate federal habeas corpus jurisdiction. Nor does section 16-1901 include a “safety valve” such as provided for in séction 23-110, strongly suggesting that Congress did not think that section 16-1901 raised the same constitutional issues that it foresaw in section 23-110. Yet without a safety-valve, there would indeed be serious questions as to whether section 16-1901 violates the Constitution’s Suspension Clause if it is interpreted to foreclose any access to an Article III court however “inadequate or ineffective” the local court proceedings. See Swam, 430 U.S. -at 381, 97 S.Ct. 1224 (noting that the safety-valve “avoids any serious question about the constitutionality” of section 23-110); see also Weaver v. United States Information Agency, 87 F.3d 1429, 1436 (D.C.Cir.1996), cert. denied, - U.S. -, 117 S.Ct. 2407, 138 L.Ed.2d 174 (1997) (“A statute must be construed, if fairly possible, so as to avoid not only the conclusion that it is unconstitutional but also grave doubts upon that score.”).
The most natural construction of sections 16 — 1901(b) and (c) is that they are venue provisions applicable only to petitions for ha-beas corpus made under section 16-1901 itself. Sections (b) and (c) both set forth the courts in which “petitions for writs” may be filed, but do not specify what types of writs are meant. Given that the immediately preceding section, 16-1901(a), grants authority for persons held within the District of Columbia to petition for a writ of habeas corpus, it is appropriate to assume' that subsections (b) and (c) refer to writs under that section. Broader alternative interpretations of those two subsections bring in their wake a host of problems. It seems unlikely for instance that those subsections are intended to refer to any kind of petitions for writs, including petitions for writs of mandamus, prohibition, and so forth. Nor is there any obvious textual reason to read subsections (b) and (c) to reach some types of petitions other than those made under section 16-1901(a), but not others — for instance, petitions under sections 2241 and 2254, but not mandamus petitions.
3. The history of section 16-1901
, The limited available evidence as to the history of section 16-1901 confirms — or at least fails to contradict — our view that sections 16-1901 and 2241 are properly conceived of as distinct, equally available avenues by which D.C. petitioners may seek habeas corpus. A version of section 16-1901 existed long before the DCCRCPA; indeed, a predecessor was enacted in the course of Congress’s 1901 recodification of the D.C.Code. See 31 Stat. 1372 (1901). The earlier section largely tracked the language of what is now section 16-1901(a), but with a reference to “the United States District Court for the'District .of Columbia” where that section now refers to “the appropriate court'.” Before the enactment of the DCCRCPA, section 2241 contained no territorial restriction that would have precluded its invocation by petitioners within the District of Columbia; it permitted (and still permits) “the Supreme Court, any justice thereof, the district courts and any circuit judge” to grant writs of habeas corpus “within their respective jurisdictions.” 28 U.S.C. § 2241. Nor did the former text of section 16-1901 say that it was the exclusive remedy for D.C. habeas corpus petitioners. Although the caselaw does not explain the relationship between the two sections • in any detail, it suggests that petitioners could choose to proceed under either or under both. See, e.g., Stewart v. Overholser, 186 F.2d 339, 340 (D.C.Cir.1950) (as to a prisoner at St. Elizabeths, citing the general federal habeas corpus statute, 28 U.S.C. § 2241 et *1045seq., and saying that the court has “considered also the statutes of the District of Columbia [citing what is now § 16-1901],” and found, “[ajssuming them to be in effect,” that they “do not require a different result”); Dorsey v. Gill, 148 F.2d 857, 867 (D.C.Cir.1945) (describing the requirements of what are now§ 16-1901 and 28 U.S.C. § 2241 et seq. as “equivalent and synonymous”).
Our conclusion that before the DCCRCPA was enacted petitioners could ordinarily choose to proceed under either or both of the two provisions is strengthened by section 16-1909 of the D.C.Code (enacted in 1968), which says: “This chapter [§§ 16-1901 — 16-1909] does not affect any provision of chapter 153 [Habeas. Corpus] of title 28, United States Code.” The accompanying Revision Note states:
Section is new, and is inserted for the purpose of construction.
Chapter 153 of Title 28, United States Code, also relates to habeas corpus and applies to Federal courts generally, including the United States District Court for the District of Columbia. Upon the reenactment of the provisions carried into this chapter, they will constitute a later enactment than Title 28, United States Code, which was enacted in 1948. Considering the local character of the provisions carried into this chapter, there should not arise, as a general rule, even without this section, any question of conflict. However, this section is inserted as a precautionary measure.
Chapter 153, of course, includes section -2241 and the other general federal habeas corpus provisions. Congress’s express reference to “the United States District Court for the District of Columbia” indicates that Congress intended that section 2241 should continue to remain available in the federal courts in the District, despite the existence of section 16-1901.7 See also McCall, 510 F.2d at 182 (saying that section 16-1909 “specifies that the District habeas provisions are not to be construed to restrict the general federal ha-beas provisions”).
But if sections 16-1901(b) and (e) only set forth the venue for those habeas corpus petitions filed under section 16-1901(a), why— one may ask — the reference to petitions “directed to Federal officers and employees” in sectionl6-1901(b)? Here again, history is of some help. At the time the DCCRCPA was enacted, there were a large number of persons in the D.C. prisons who had been tried in federal district court. As we explained in McCall, subsection (b) ensured that any petitions that they filed under section 16-1901 would be heard in federal court. This is so because the local prison officials who were their custodians were “deemed to be ‘Federal officers or employees’ within the meaning of 16 D.C.Code § 1901” because they had custody over a federal prisoner. McCall, 510 F.2d at 180. When a prisoner is convicted and sentenced by the Superior Court, this logic does not apply, and exclusive jurisdiction over actions under section 16-1901 is vested in the Superior Court.8 Because sections 16-1901(b) and (c) serve this narrower function, there is no need to read them as broadly as the District proposes, to allocate jurisdiction under all habeas corpus statutes between the federal and D.C. courts.
Portions of the legislative history of the DCCRCPA do suggest that Congress thought that the federal courts would not be able to entertain any habeas corpus petitions *1046brought by persons convicted in the newly-created D.C. courts. For example, the report of the Senate Committee on the District of Columbia said that after the effective .date of the DCCRCPA, “the U.S. District Court for the District of Columbia shall no longer have jurisdiction over local civil actions in the nature of quo warrento, suits to acquire title, and actions regarding change of name, contractors’ bonds, restitution of real property, replevin of personal property, habeas corpus, and commitment of narcotics users.... ”S. Rep. No. 91-405, at 19 (1969); see also id. at 23; H.R. Rep. No. .91-901 at 136 (1970), U.S. Code Cong. & Admin. News at 2966 (saying that as of the effective date of the Act the Superior Court will have jurisdiction over a range of matters, including “writs of habeas corpus to persons other than Federal officers and employees”); Staff of the Senate Committee ON THE DISTRICT OF COLUMBIA, 91ST Cong., Statement of the Managers On the part of the Senate 5-6 (Comm. Print 1970) (saying that the Superior Court is to have “exclusive, general jurisdiction over all local matters, civil and criminal,” and listing “ha-beas corpus” as an example). The DCCRCPA’s legislative history, however, almost universally refers to “habeas corpus” generally,- without distinguishing those petitions that-attack the petitioner’s conviction or sentence and are given over exclusively to the local courts under section 23-110(g) from other habeas petitions which that section does not encompass. We confess that we can discern no obvious reason for Congress to deny the federal courts jurisdiction over one type of petition but permit jurisdiction over the other. Congress’s description of “habeas corpus” petitions as “local” in character, S. Rep. No. 91-405, at 19 (1969), seems to fit both types of petition equally well (with one possible exception, discussed below).
But, because Congress did not explicitly rémove our jurisdiction over those section 2241 actions that do not attack the petitioner’s conviction or sentence, that jurisdiction necessarily continues in effect. This is plain from Felker v. Turpin, 518 U.S. 651, 116 S.Ct. 2333, 135 L.Ed.2d 827 (1996), in which the Supreme Court found that although the AEDPA bars habeas corpus petitioners who are denied leave to file a second or successive petition from challenging this decision through certiorari, see AEDPA § 106(b)(3)(E), it does not withdraw the Court’s, authority to entertain original habeas corpus petitions under 28 U.S.C. .§§ 2241 and 2254. This was so even though a petitioner may be able to use the latter route to circumvent Congress’s apparent intent in enacting the AEDPA to bar all review of such decisions. , The rule that “[rjepeals by implication are not favored,” Felker, 116 S.Ct. at 2338, and the principle that the withdrawal of habeas corpus jurisdiction is subject to especial scrutiny, dates back over a hundred years; in Ex parte Yerger, 75 U.S. (8 Wall.) 85, 19 L.Ed. 332 (1868), the Court similarly declined to conclude that Congress had implicitly stripped it of jurisdiction to hear ha-beas corpus actions under the forerunner of section 2241.
Congress’s intent in enacting the DCCRCPA is if anything less clear than it was in Felker. This is so because reading sections 16-1901(b) and (c) to bar us from exercising jurisdiction over all D.C. habeas petitions that are not directed at federal officials, as the District urges, might yield anomalous results in some cases. For example, Crowell, the petitioner in one of the companion cases, was convicted in the state courts of Virginia but is now being held in D.C.’s prison system. His habeas corpus petition must be directed to his custodian, who is either a D.C. official, or perhaps (under the principle of McCall) constructively an official of Virginia. Either way, Crowell’s custodian is not a federal official, so that, under the District of Columbia’s proposed reading of section 16-1901, exclusive jurisdiction over Crowell’s petition would be in D.C. Superior Court. Sée § 16-1901(c) (“Petitions for writs directed to any other person shall be filed in the Superior Court of the District of Columbia.”). Yet, had Crowell been transferred to a prison anywhere else in the nation,- he would have had a remedy in an Article III court. Cf. McCall, 510 F.2d at 182 (citing the anomaly of denying access to a federal habeas corpus remedy to only those federal prisoners convicted in the District of Columbia in concluding that section 16-1901 permits such prisoners to file their petitions *1047in federal court). And Crowell’s action, which involves the good time credit rules of the Commonwealth of Virginia, does not seem to be the type of “local civil action” that Congress had in mind when it enacted section 16-1901. See S. Rep. No. 91^405, at 19 (1969) (saying that the DCCRCPA would give the Superior Court exclusive jurisdiction over all “local civil actions,” including proceedings in habeas corpus). In short, not only does the text of section 16-1901-fail to divest the federal courts of jurisdiction under section 2241, but the underlying intent of Congress in enacting that provision is far from clear. Thus, we now find that Congress has not extinguished Blair-Bey’s section 2241 remedy.
C. The Merits of Blair-Bey’s Action
In one of the companion cases, Crowell v. Walsh, No. 97-7192, we, conclude that cases filed before the AEDPA was enacted are governed by the preexisting certificate of probable cause requirement, not by the new certificate of appealability requirement. Blair-Bey’s action was filed before the AED-PA was enacted; thus, we will consider whether he is entitled to a certificate of probable cause under our pre-AEDPA standards.
Blair-Bey makes claims under the Due Process Clause and under the Ex Post Facto Clause. Blair-Bey is not entitled to a certificate of probable cause as to his due process claim, because it is painfully clear that he cannot make any successful claim of Fifth Amendment violation. As to his ex post facto claim, however, we grant the certificate of probable cause, and remand to permit Blair-Bey to develop his claim further.
1. Due Process Clause
Blair-Bey cannot make out a due process claim because he cannot point to a constitutionally protected, liberty interest of which he has been deprived without due process. Liberty interests are of two types: those issuing directly from the Constitution and those created-by state law. Blair-Bey does' not, of course, have a -direct constitutional liberty interest in parole. Greenholtz v. Inmates of Nebraska Penal and Correctional Complex, 442 U.S. 1, 99 S.Ct. 2100, 60 L.Ed.2d 668 (1979). As to liberty interests created by state law, mandatory language in applicable state laws and regulations may suffice to create a liberty interest. See, e.g., Heuntt v. Helms, 459 U.S. 460, 471-72, 103 S.Ct. 864, 74 L.Ed.2d 675 (1983).9 But Blair-Bey cannot point to any such mandatory language here; the applicable D.C. parole regulations say only that “reconsideration shall ordinarily occur within twelve (12) months.” D.C. Mun. Regs. tit. 28, § 104.2, 35 D.C.Reg. 455 (1988) (emphasis added), and add that “[njotwithstanding any other provision of this section, the Board may order a parole reconsideration date it determines to be appropriate.” Id., § 104.11. The D.C. Court of Appeals has found the applicable regulations not to create a liberty interest; see Stevens v. Quick, 678 A.2d 28, 31-32 (D.C.1996) (a case also involving a five-year setoff), and we agree.
Blair-Bey does point to a set of guidelines established by the Parole Board to guide its set-off decisions. The guidelines do not appear in the record; they are, however, quoted extensively in the D.C. Court of Appeals’s Hall v. Henderson opinion, 672 A.2d 1047 (1996), which also involved a challenge by a D.C. inmate to a five-year set-off.10 The guidelines list a series of “aggravating” and “mitigating” factors for the Board to consider in making set-off decisions. The D.C. Court of Appeals observed that the guidelines do require the DCBOP to have “some” basis for deviating from the normal set-off period. Nevertheless, the court found that the guide*1048lines-do not create a liberty interest, because they do not limit which factors the DCBOP can consider, or how to weigh them. Id- at 1054. We agree that so discretionary and open-ended a document cannot be construed to give rise to a liberty interest.11
2. Ex Post Facto Clause
Blair-Bey has a more substantial claim that the rules governing the District’s parole system were made stricter after he committed his crimes, and that the present rules as applied to his case amount to a prohibited ex post facto law. D.C. law provides:
Whenever it shall appear to the-Board of Parole that there is a reasonable probability that a prisoner will live and remain at liberty without violating the law, that his release is not incompatible with the welfare of society, and that he has served the minimum sentence imposed or the prescribed portion of his sentence, as the case may be, the Board may authorize his release on parole upon such terms and conditions as the Board shall from time to time prescribe.
D.C.Code § 24-204 (1981). Before 1987, the DCBOP’s implementing regulations simply mirrored this provision, requiring only that in exercising its discretion the Board consider a list of factors, including the inmate’s offense, prior history, of criminality, personal and social history, physical and emotional health, institutional experience, and availability of community resources. 9 D.C.R.R. § 105.1(a)-(f) (1981). -
In 1987, the Board of Parole promulgated new regulations that provided for the use of “salient factor scores” in making parole determinations. This scheme takes into account factors much like those considered under the preexisting regulations, but provides a scoring system for weighing those factors. Once a score has been calculated for a particular inmate, the new regulations indicate whether parole should ordinarily be granted or denied. D.C. Mun. Regs., tit. 28, § 204.1, 204.4-204.18. Under the regulations, the Board still retains discretion in making individualized parole determinations; the scoring system is intended “to guide-the Board in making the • decision whether to grant or deny parole.” Davis v. Henderson, 652 A.2d 634, 635 (D.C.1995) (describing the transition from the old to the new parole system).
Blair-Bey contends that under the new salient factor scoring system his criminal history will always produce a score that invokes á presumptive denial of parole, irrespective of any rehabilitation he may undergo in prison. He asserts that he would have fared better under the more open-ended parole system that was in effect in 1975 and 1980 (when he committed the two murders that led to his incarceration): under the earlier regulations he would not have been subject to any presumptive denial of parole. He says that subjecting him to the post-1987 parole rules amounts to imposing an ex post facto law in his ease, in violation of Article I, § 10 of the Constitution (“No State shall ... pass any ... ex post facto Law ... ”).
The constitutional bar on the enactment of ex post facto laws means that “[ljegislatures may not retroactively alter the definition of crimes or increase the punishment for criminal acts.” Collins v. Youngblood, 497 U.S. 37, 43, 110 S.Ct. 2715, 111 L.Ed.2d 30 (1990). Blair-Bey argues that the 1987 adjustment to the District of Columbia’s parole system increases the “punishment” attached to his crime.
The circuit caselaw poses something of a problem for Blair-Bey. In Warren v. United States Parole Commission, 659 F.2d 183 *1049(D.C.Cir.1981), we rejected a claim by a prison inmate that the 1976 revision of the federal parole system was an ex post facto law. Before 1976, the federal parole system had also been highly discretionary; the 1976 revision established a “salient factor score” system that structured this exercise of discretion. The 1976 revision thus effected a change much like (but, as we observe below, not necessarily exactly like) the 1987 revision in the D.C. parole system that Blair-Bey protests. Warren concluded that the 1976 revision could not be said to increase the “punishment” of prospective parolees, and hence was not an ex post facto law. Warren reasoned that the Parole Commission retained discretion to ignore its own guidelines; that the guidelines had been based on a statistical survey of past parole practice, so that in the aggregate they “embody what may well have been the Board’s practice anyway,” id. at 193 (emphasis omitted); that under the old system it would always be difficult to predict how any particular inmate would have been treated; and finally that, at most, the new parole system reduced the likelihood that an inmate would be paroled either much earlier or much later than average, a change that could not be characterized as “worsening]” the inmate’s position. Id. at 193-94. See also Miller v. Florida, 482 U.S. 423, 434, 107 S.Ct. 2446, 96 L.Ed.2d 351 (1987) (in the course of deciding another case, seemingly assuming that Warren and similar eases in other courts of appeals upholding the 1976 revisions in the parole system were correctly decided).
Because the district court dismissed Biair-Bey’s petition sua sponte, Blair-Bey had no opportunity to develop arguments and present evidence below. It is therefore impossible to determine whether or not his claim falls under the Warre-wrationale. In particular, we do not know what evidence Blair-Bey and the DCBOP might present as to the purpose and effect of the 1987 revision of the D.C. parole system. If the 1987 revision undertook to codify past parole practices, in the way that the 1976 federal revision did, and if the DCBOP in practice retains discretion to ignore the guidelines, then his case will fall under Warren. But Blair-Bey may be able to present evidence distinguishing his case from Warren, in any of three ways. First, he may be able to show that the revisions to the DCBOP scheme impose a sufficiently great risk of disadvantaging a particular category of inmates as to violate the ex post facto clause. See California Department of Corrections v. Morales, 514 U.S. 499, 509, 115 S.Ct. 1597, 131 L.Ed.2d 588 (1995) (declining to decide what degree of risk is “sufficient,” but finding that the “speculative and attenuated” risk of harm in the ease at bar did not suffice). Second, if he can show that the 1987 revision was motivated by a punitive desire to extend the incarceration of a particular category of inmates, see Miller, 482 U.S. at 433-34, 107 S.Ct. 2446 (finding that a statute whose “sole reason” was to “punish sex offenders more heavily” violated the ex post facto clause), his case will be strengthened. And third, if the DCBOP does not in practice ever ignore its own guidelines, this too may lead to the conclusion that the broad discretion relied upon in Warren is absent here. Cf. Warren, 659 F.2d at 197 n. 57 (leaving open the question of whether evidence that the Board generally engaged in a “mechanical administrative application .of the Guidelines ... could implicate the ex post facto clause”). One of these factors, or a combination of them, could suffice to establish that Blair-Bey has been deprived of his “entitlement to have [the Parole Board’s] discretion exercised,” id. at 196; we need not decide now what kind of showing would suffice.12
The D.C. Court' of Appeals has already rejected an ex post facto challenge to the 1987 revisions to the D.C. parole system. See Davis, 652 A.2d at 636. Davis found no ex post facto violation because “[t]he new District of Columbia parole guidelines ... *1050merely formalize the manner in which..the Board exercises the- discretion conferred upon it by the governing provision in effect when Davis was sentenced.” Id. We are bound to follow interpretations of D.C. law by the D.C. Court of Appeals, and hence must defer to that court’s ruling to the extent that it interprets D.C. law; for instance, we defer to the ruling that the D.C. parole guidelines “merely formalize the manner in which the Board” exercises its discretion. But we are not bound to follow the D.C. Court of Appeals’s analysis of federal law, and so need not defer to Davis’s reading of what the ex post facto clause requires on the facts of a particular case. Moreover, it appears that the plaintiff in Davis did not raise any of the grounds that we. have identified as potentially distinguishing- this case from Warren. Thus, should Blair-Bey present evidence on any of these points, the specific holding of Davis would not necessarily control even the district court’s interpretation of D.C. law (although Davis’s methodology would of course guide its analysis).
III. Conclusion
In sum, we conclude that this ease is not subject to the filing-fee rules of the PLRA, and that the district court erred in finding that D.C.Code section 16-1901 barred the federal courts from considering Blair-Bey’s habeas corpus petition. We also find that the United States Parole Commission should be dismissed as a party to this action. On the merits, we find that Blair-Bey’s due process claim is without foundation. However, because it is possible that Blair-Bey may be able to make out an ex post facto claim, we remand that part of his case for reconsideration in accordance with the principles set forth in this opinion.
So ordered.
. Blair-Bey was confined at Occoquan when he filed his petition; while this appeal was pending, he was transferred to a prison in Ohio. "Notwithstanding [Blair-Bey]'s transfer during the course of this litigation" to the Ohio prison, "habeas jurisdiction as a general matter continues to be in the district where the prisoner was incarcerated at the time the habeas petition was filed.” See id. at 806 n. 1.
. The word "appeal" iA the statute does not reach all appeals, but only appeals in civil actions. See United States v. Simmonds, ill F.3d 737, 744 (10th Cir.1997); Martin v. United States, 96 F.3d 853, 854 (7th Cir.1996).
. District notes that Senator Dole cited a study of the increasing number of "due process and cruel and unusual punishment” complaints filed by prisoners, 141 Cong. Rec. S7498, S7524 (daily ed. May 25, 1995), and argues that some parole-related habeas petitions are based on the due process clause (indeed, it is one of the grounds for Blair-Bey’s petition). But an article by the author of that very study was appended to the floor statement of Senator Kyi; it reveals that the study focused on due process challenges to violations of prison regulations, not parole regulations. See Walter Berns, Sue the Warden, Wall Street Journal (Apr. 24, 1995) (reprinted in 141 Cong. Rec. S7498, S7527 (daily ed. May 25, 1995) (statement of Senator Kyi)).
. That is, an inmate unable to pay the $5 filing fee for habeas petitions (or, if the inmate loses in district court, the more daunting $100 fee for filing an appeal).
. By contrast, the AEDPA's comparable repeat-filer provision is not nearly as restrictive. Its bar on successive habeas corpus petitions would only apply if an inmate attempted to challenge the same official act twice. As Newlin observed, " [successive challenges to the same parole decision or revocation of good-time credits are rare if not nonexistent.” Newlin, 123 F.3d at 438.
. Its relationship to D.C.Code § 23-110 is thus much like the relationship, for federal prisoners, of 28 U.S.C. § 2241 and 28 U.S.C. § 2255: the former provides a broad habeas corpus remedy, the latter a specific instrument for attacking a conviction or sentence.
. The USPC argues that section 16-1909 is only intended to make clear that prisoners convicted in the District and then transferred to prisons elsewhere are not within section 16-1901. But this is already obvious from the text of section 16-1901 itself (which only applies to prisoners “in the District”). This reading is also inconsistent with the Revision Note.
. Sections (b) and (c) also have the effect of preventing the courts of the District of Columbia from entertaining section 16-1901 petitions brought against federal officials.
The USPC cites to isolated passages of McCall which it claims support its reading of section 16-1901. For example, it points to a statement in a footnote in which, after outlining the structure of section 16-1901 and its allocation of authority between the Superior Court and federal district court, the court said: "Thus habeas jurisdiction within the District of Columbia is exclusively vested in one court or the other." 510 F.2d at 170 n. 3. In context, it is clear that this dictum was a reference only to habeas corpus jurisdiction under section 16-1901 itself, not necessarily to jurisdiction under some other statute.
. In Sandin v. Conner, 515 U.S. 472, 115 S.Ct. 2293, 132 L.Ed.2d 418 (1995), the Supreme Court adjusted the Hewitt analysis in considering a prisoner’s challenge of his placement in disciplinary segregation. In Ellis v. District of Columbia, 84 F.3d 1413 (D.C.Cir.1996), we found that SandinotAy alters the liberty-interest analysis applicable to claims relating to "the day-to-day management of prisons,” and that it does not , apply to parole-related claims. See id. at 1418. Thus, Blair-Bey’s action is not subject to Sandin.
. We assume that the samé set of guidelines was applied to Blair-^Bey. Although we have no way to be sure of this,' he has not suggested that the applicable guidelines have changed.
. Blair-Bey also argues that the Board’s conduct was so arbitrary as to violate the due process clause even in the absence of an identifiable liberty interest. ' There is some authority for the proposition' that exceptionally arbitrary governmental conduct may in itself violate the due process clause, whether or not a liberty or property interest is at. stake. .See, e.g., Burkett v. Love, 89 F.3d 135, 139-740- (3d Cir.1996) (denial of parole in retaliation for an inmate’s successful habeas petition); see also Perry v. Sindermann, 408 U.S. 593, 597, 92 S.Ct. 2694, 33 L.Ed.2d 570 (1972). But the record in this case is devoid of any evidence that might meet so high a standard. Blair-Bey asserts that the Board relied on his juvenile record in making its decision, and claims that this violated applicable municipal regulations. Even supposing that this is.true (a point that is far from clear), such reliance would not be so irrational or arbitrary as to violate the due process clause.
. There is also a question whether parole regulations count as "laws" at all for purposes of the constitutional prohibition on "ex post facto laws." Compare Flemming v. Oregon Board of Parole, 998 F.2d 721 (9th Cir.1993) (Oregon’s parole regulations are “laws” for ex post facto purposes) with Bailey v. Gardebring, 940 F.2d 1150, 1156 (8th Cir.1991) (Minnesota’s are not); see also Bailey, 940 F.2d at 1157 (reviewing cases addressing this issue). Wanwideclined to resolve this issue, see 659 F.2d at 197 n. 57, and so do we at so early a stage in the litigation.