with whom Mr. Justice Brennan and Mr. Justice Stewart* join, dissenting.
I
Mr. Justice Blackmun’s plurality opinion, by its efforts to establish that O’Callahan v. Parker, 395 U. S. 258 (1969), was not a decision dealing with jurisdiction in its classic form, implicitly acknowledges that if O’Callahan were in fact concerned with the adjudicatory *694power — that is, the jurisdictional competency1 — of military tribunals, its holding would necessarily be fully retroactive in effect, cf. e. g., Linkletter v. Walker, 381 U. S. 618, 623 (1965). The plurality now puts forth the view that O’Callahan was not concerned with the true jurisdictional competency of courts-martial but that the decision yielded merely a new constitutional rule. This characterization of O’Callahan permits the plurality to apply in this case the three-prong test employed to judge the retroactivity of new procedural rules under Linkletter and its progeny, see, e. g., Desist v. United States, 394 U. S. 244, 249 (1969); Stovall v. Denno, 388 U. S. 293, 297 (1967). And, not surprisingly, application of that test leads to the conclusion that O’Callahan should have only prospective effect. With all due respect, I must dissent.
I am unable to agree with the plurality’s characterization of O’Callahan. In my view, it can only be understood as a decision dealing with the constitutional limits of the military’s adjudicatory power over offenses committed by servicemen. No decision could more plainly involve the limits of a tribunal’s power to exercise jurisdiction over particular offenses and thus more clearly demand retroactive application.
A
In holding that O’Callahan is to be given only prospective effect, the plurality does not reject outright the view that the decision was jurisdictional in nature. Yet it clearly does reject the contention that O’Callahan dealt with a question of true jurisdictional competency, for we are told that the decision “did announce a new constitutional principle,” ante, at 673, and that it really “dealt with the appropriate exercise of jurisdiction *695by military tribunals,” ante, at 674. The difference between a decision concerning a tribunal’s jurisdictional competency — that is, the limits of its adjudicatory power — and “the appropriate exercise of [its] jurisdiction” is less than clear to me, at least where, as here, the question of “appropriateness” ultimately turns on the extent of Congress’ constitutional authority under Art. I, § 8, cl. 14, to “make Rules for the Government and Regulation of the land and naval Forces.” But whatever the nature of the distinction that the plurality now seeks to draw, it cannot, in my opinion, obscure the essential character of the decision in O’Callahan.
O’Callahan required this Court to define the class of offenses committed by servicemen that Congress, under Art. I, § 8, cl. 14, could constitutionally empower military tribunals to try. The nature of the ultimate inquiry there is plain from the question upon which the Court granted certiorari: “ ‘Does a court-martial, held under the Articles of War, Tit. 10, U. S. C. § 801 et seg.,, have jurisdiction to try a member of the Armed Forces who is charged with commission of a crime cognizable in a civilian court and having no military significance, alleged to have been committed off-post and while on leave, thus depriving him of his constitutional rights to indictment by grand jury and trial by a petit jury in a civilian court?’ ” 395 IT. S., at 261. The O’Callahan Court’s discussion of this issue was consistently couched in terms of the jurisdiction of military tribunals; 2 and, in dissent, Mr. Justice Harlan, too, framed the issue presented in the unmistakable terms of “the appropriate subject-matter jurisdiction of courts-martial,” id., at 276. Even the Court of Appeals in No. 71-6314, while ultimately holding the O’Callahan decision to be prospective only, acknowledged that the decision turned upon a determination of “lack of adjudicatory power” — that “O’Cal*696lahan’s foundation, framework and structure deny to the legislation which breathed the breath of judicial life into the forum that tried Sgt. O’Callahan, the necessary basis in constitutional power to reach his type of case.” 3 450 F. 2d 753, 757 (CA5 1071). See also United States ex rel. Flemings v. Chafee, 458 F. 2d 544, 549-550 (CA2 1972).
Despite the evident jurisdictional nature of the ultimate issue presented in O’Callahan, the plurality attempts to analogize this case to DeStefano v. Woods, 392 U. S. 631 (1968), where the Court held that the decisions in Duncan v. Louisiana, 391 U. S. 145 (1968), and Bloom v. Illinois, 391 U. S. 194 (1968), were to have only prospective effect. Duncan held that the Sixth Amendment guarantee of trial by jury in criminal cases had been made applicable to the States by the Fourteenth Amendment. And Bloom established the right to jury trial in the context of serious criminal contempt proceedings. DeStefano — like the other offspring of Linkletter that have applied the three-prong test to determine retro-activity — involved constitutional rulings that established new procedures for the conduct of trial or for the use of evidence. But O’Callahan hardly was such a case.
The Court in O’Callahan was not setting forth procedures which the military was constitutionally required to adopt in its proceedings. Had the Court been doing so, this would certainly be a different case; the analogy *697to DeStefano then might well be appropriate. It is true, as the plurality now points out, that the O’Callahan Court placed considerable emphasis on the lack of jury trial in the court-martial system. But it did so only as a part of the general analytic process of determining the proper reconciliation of the competing jurisdictions of two essentially distinct4 judicial systems, namely, the civil and military systems of justice. The Court’s basic concern in this process was the preservation — to the fullest extent possible consistent with the legitimate needs of the military — of the fundamental civil rights guaranteed by our Constitution and Bill of Rights. Those civil rights were, in the Court’s words, the “constitutional stakes in the . . . litigation.” O’Callahan v. Parker, supra, at 262.
Thus, the Court pointed out that one tried before a military tribunal is without the benefit of not only trial by jury but also indictment by a grand jury. Ibid. Nor are the same rules of evidence and procedure applicable in a military proceeding, a factor affecting, for example, the defense’s access to compulsory process, id., at 264 n. 4. In addition, the Court was concerned with the fact that the presiding officers at courts-martial do not enjoy the independence that is thought to flow from life tenure and undiminishable salary. To the contrary, the Court recognized that “the possibility of influence on the actions of the court-martial by the officer who convenes it, selects its members and the counsel on both sides, and who usually has direct command authority over its members is a pervasive one in military law, despite strenuous efforts to eliminate the danger.” Id., at 264. In short, the Court con-*698eluded that “[a] court-martial is not yet an independent instrument of justice but remains to a significant degree a specialized part of the overall mechanism by which military discipline is preserved,” Id., at 265.
The Court’s purpose in considering these factors was not to require changes in the military system of justice, but rather to illustrate its “fundamental differences from . . . the civilian courts,” id., at 262, differences that compelled the Court “ 'to restrict military tribunals to the narrowest jurisdiction deemed absolutely essential to maintaining discipline among troops in active service,’ ” id., at 265, quoting from Toth v. Quarles, 350 U. S. 11, 22 (1955). As a result, the Court concluded that the “crime to be under military jurisdiction must be service connected . . . ,” 395 U. S., at 272, so that the power of Congress under Art. I, § 8, cl. 14, to “make Rules for the Government and Regulation of the land and naval Forces,” and also the exemption from the grand jury requirement of the Fifth Amendment for “cases arising in the land or naval forces, or in the Militia, when in actual service in time of War or public danger” are not expanded to deprive servicemen unjustifiably of their civil rights.5 The Court found that when an offense is not service *699connected, the needs of the military are not significantly implicated and thus that the limits of Congress' constitutional power over servicemen under Art. I, § 8, cl. 14, have been passed, at least in the context of “peacetime offenses,” 395 U. S., at 273.
Certainly the jurisdictional nature of the O’Callahan decision is amply demonstrated by this Court’s previous decision in McClaughry v. Deming, 186 U. S. 49 (1902). There the Court was called upon to decide “the power of an officer convening a court-martial for the trial of an officer of volunteers [reserve troops], to compose that court entirely of officers of the Regular Army.” Id., at 53. The Court determined that Congress had directed by statute that volunteer officers of the Army be tried only by a court-martial composed of volunteer officers. In light of this determination the Court concluded:
“As to the officer to be tried there was no court, for it seems to us that it cannot be contended that men, not one of whom is authorized by law to sit, but on the contrary all of whom are forbidden to sit, can constitute a legal court-martial because detailed to act as such court by an officer who in making such detail acted contrary to and in complete violation of law. Where does such a court obtain jurisdiction to perform a single official function? How does it get jurisdiction over any subject-matter or over the person of any individual? The particular tribunal is a mere creature of the statute, as we have said, and must be created under its provisions.” Id., at 64.
In the same vein, the Court elsewhere stated: “A court-martial is the creature of statute, and, as a body or tribunal, it must be convened and constituted in entire conformity with the provisions of the statute, or else it is without jurisdiction.” Id., at 62. Because of the flaw *700in the composition of the court-martial, a flaw which the Court considered determinative on the issue of the court-martial’s jurisdiction, the Court affirmed a lower court’s issuance of a writ of habeas corpus to secure the officer’s release from military custody. Significantly, this writ was issued at a time when habeas corpus clearly lay only where the court-martial had “no jurisdiction over the person of the defendant or the subject-matter of the charges against him.” Id,., at 69.6 In O’Callahan the Court was not concerned with the composition of a particular court-martial, but with the fundamental question of the extent of Congress’ constitutional power to establish court-martial jurisdiction over offenses committed by our servicemen. If the former issue goes to the jurisdiction of military tribunals, certainly the latter does.
B
With this understanding of O’Callahan, I believe, contrary to the plurality’s view, that the retroactive application of our holding there is required by our prior decisions in Robinson v. Neil, 409 U. S. 505 (1973), and United States v. U. S. Coin & Currency, 401 U. S. 715, 722-724 (1971). Robinson involved the retroactive application of the decision in Waller v. Florida, 397 U. S. 387 (1970), that the Fifth Amendment’s guarantee, made applicable to the States through the Fourteenth Amendment, that no person should be put twice in jeopardy for the same offense barred an individual’s prosecution for a single offense by both a State and a municipality of the State, that is, a legal subdivision of the State. U. S. Coin & Currency held retro*701active the Court’s prior determination that the Fifth Amendment privilege against compulsory self-incrimination barred the prosecution of gamblers for failure to register and to report illegal gambling proceeds for tax purposes, see Marchetti v. United States, 390 U. S. 39 (1968); Grosso v. United States, 390 U. S. 62 (1968).
In deciding whether to give retroactive effect to Waller, Marchetti, and Grosso, the Court rejected contentions that it should apply the three-prong test employed in cases such as Stovall v. Denno, 388 U. S. 293 (1967), Desist v. United States, 394 U. S. 244 (1969), and DeStefano v. Woods, 392 U. S. 631 (1968). In U. S. Coin & Currency, Mr. Justice Harlan, speaking for the Court, explained:
“Unlike some of our earlier retroactivity decisions, we are not here concerned with the implementation of a procedural rule which does not undermine the basic accuracy of the factfinding process at trial. Linkletter v. Walker, 381 U. S. 618 (1965); Tehan v. Shott, 382 U. S. 406 (1966); Johnson v. New Jersey, 384 U. S. 719 (1966); Stovall v. Denno, 388 U. S. 293 (1967). Rather, Marchetti and Grosso dealt with the kind of conduct that cannot constitutionally be punished in the first instance.” 401 U. S., at 723.
The Robinson Court adopted essentially the same view of the Waller decision concerning the Double Jeopardy Clause and multiple prosecutions by different legal subdivisions of a single sovereign. See 409 U. S., at 508. In this case, too, we are concerned, not with “the implementation of a procedural rule,” but with an unavoidable constitutional impediment to the prosecution of particular conduct.
In O’Callahan, as has been seen, the ultimate issue was the extent of the constitutional power that underlies *702the jurisdiction of military tribunals. Where an offense lies outside the limits of that power, there exists just as much of a constitutional impediment to trial by court-martial as there existed to a civilian trial in Marchetti and Grosso due to the privilege against self-incrimination or in Waller due to the Double Jeopardy Clause. It cannot be forgotten that military tribunals are courts of limited jurisdiction. See McClaughry v. Deming, 186 U. S., at 63; Ex parte Watkins, 3 Pet. 193, 209 (1830). They cannot exercise authority which Congress has not conferred upon them, much less authority which Congress is without constitutional power to confer.7 It is this fundamental principle that compels retroactive application of the decision in O’Callahan.
The plurality seeks to distinguish U. S. Coin & Currency and Robinson on the grounds that the former involved a right that prevented the offender from being tried at all and the latter a right that prevented "another trial from taking place at all,” ante, at 679, whereas the underlying issue in this case is merely which jurisdiction can try offenses committed by servicemen. But these are distinctions without meaning; they *703merely reflect the differences in the nature of the constitutional impediment to trial at issue in each case. The essential common thread tying these cases together is that each involved, at the least, a constitutional barrier to trial before the particular forum, regardless of the fairness of the procedures and the factfinding process of the relevant forum.
U. S. Coin & Currency swept broadly, to be sure, for it concerned a constitutional guarantee that effectively prevented any trial of the offender for the particular offense. But the nature of the Double Jeopardy Clause at issue in Robinson is such that the offender may be tried once for a particular offense by a court of a particular sovereign; it is the second prosecution for the same offense by another court of the same sovereign that that Clause clearly bars. Similarly here, a serviceman charged with a nonservice-connected offense is subject to trial for that offense by civil tribunals, but military tribunals lack the necessary constitutional power, at least in peacetime, to try such an offense. As was true in Robinson, this case involves a constitutional barrier to adjudication of a particular offense by a particular forum, yet in neither case does it follow that the offender is constitutionally entitled to go unpunished altogether. I fail to see, therefore, why different rules from those applied only recently in Robinson should be applied in this case.
There is, of course, the additional fact that the Robinson Court left open the question whether reasonable, official reliance upon a particular rule might properly be considered “in determining retroactivity of a nonpro-cedural constitutional decision such as Waller.” 409 U. S., at 511.8 And in this case the plurality, in attempt*704ing to establish that O’Callahan was a “ ‘clear break with the past/ ” ante, at 672, citing Desist v. United States, 394 U. S., at 248, and should therefore be applied only prospectively, does make much of the argument that substantial, justifiable reliance was placed on pre-O’Callahan law concerning the exercise of court-martial jurisdiction over servicemen, see ante, at 672-673. But I seriously question the relevance of any inquiry into official reliance on prior law where, as here, the issue is jurisdictional competency. Even assuming for the moment that O’Callahan completely reinterpreted the limits of Congress' power to confer jurisdiction on courts-martial, the decision involved the authoritative construction of a constitutional provision and no military tribunal could ever constitutionally have had more power than resided therein. But the real point is that O’Callahan did not mark a sharp, new departure from prior law.
The plurality acknowledges that O’Callahan did not involve the overruling of any prior precedent, ante, at 673. It is true, as the plurality indicates, that a number of prior decisions had suggested that “military status in itself was sufficient for the exercise of court-martial jurisdiction,” ibid. Yet none of the cases upon which the plurality relies dealt in fact with a nonservice-connected offense committed by a serviceman in peacetime.9 It is fair to say, in short, that until O’Callahan *705the Court had not directly faced the issue of the service-connected nature of servicemen's offenses.
More importantly, perhaps, the O’Callahan Court’s efforts to define the constitutional limits of the jurisdiction of courts-martial was hardly the beginning of such efforts by the Court. O’Callahan was but one of a series of steps taken by this Court since the conclusion of the Second World War to restrict military jurisdiction to its constitutionally appropriate limits. Thus, in Toth v. Quarles, 350 U. S. 11 (1955), the Court ruled that a discharged serviceman could not be tried by a court-martial for offenses committed while a member of the Armed Forces. Subsequently, it was established that courts-martial did not have jurisdiction to try offenses committed by civilian dependents accompanying military personnel *706serving overseas. Kinsella v. Singleton, 361 U. S. 234 (1960); Reid v. Covert, 354 U. S. 1 (1957). Finally, the Court held that civilians employed with the military overseas were not subject to court-martial jurisdiction. See Grisham v. Hagan, 361 U. S. 278 (1960); McElroy v. Guagliardo, 361 U. S. 281 (1960). This series of cases limited the reach of courts-martial to members of the Armed Forces; they did not require the Court to go on to define the breadth of offenses for which servicemen could be tried by courts-martial. Nonetheless, these cases and O’Callahan clearly were all pieces of the same cloth. Under these circumstances, I seriously doubt that retroactive application would do substantial violence to any legitimate, official reliance upon prior law10 — even assuming that to be a valid consideration here.11
*707II
Mr. Justice Douglas, in his concurring opinion, contends that petitioner Gosa’s case merits reargument to consider whether he should be denied relief because he failed to raise his jurisdictional objection before the court-martial that tried him. Mr. Justice Douglas intimates that since the jurisdiction of the military to try petitioner was not initially contested, “res judicata [may now bar] inquiry” into the question of jurisdiction, ante, at 689. In my opinion, such an argument is clearly untenable, and hence reargument of petitioner Gosa’s case is unnecessary.
A
One of the most basic principles of our jurisprudence is that subject-matter jurisdiction cannot be conferred upon a court by consent of the parties. See, e. g., American Fire & Casualty Co. v. Finn, 341 U. S. 6, 17-18 (1951); Industrial Addition Assn. v. Commissioner, 323 U. S. 310, 313 (1945); People’s Bank v. Calhoun, 102 U. S. 256, 260-261 (1880); Cutler v. Rae, 7 How. 729, 731 (1849).12 An objection to the adjudicatory power of a tribunal may generally be raised for the first time at any stage of the litigation.13 See, e. g., Flast v. Cohen, 392 U. S. 83, 88 n. 2 (1968); United States v. Griffin, 303 U. S. 226, 229 (1938); Fortier v. New Orleans National Bank, 112 U. S. 439, 444 (1884). Those principles are applicable even in the context of collateral attacks upon *708court-martial proceedings, as is evident from this Court’s decision in McClaughry v. Deming, 186 U. S. 49 (1902).
McClaughry, as previously indicated, involved a collateral attack upon the court-martial conviction of a volunteer officer who claimed that the Regular Army court-martial which had tried him had been constituted in violation of the relevant law and therefore was without jurisdiction. The volunteer officer had failed to raise this jurisdictional objection before the court-martial, and the military contended before this Court that “his consent waived the question of invalidity,” id., at 66. The Court rejected his contention, saying:
“It was not a mere consent to waive some statutory provision in his favor which, if waived, permitted the court to proceed. His consent could no more give jurisdiction to the court, either over the subject-matter or over his person, than if it had been composed of a like number of civilians .... The fundamental difficulty lies in the fact that the court was constituted in direct violation of the statute, and no consent could confer jurisdiction over the person of the defendant or over the subject-matter of the accusation, because to take such jurisdiction would constitute a plain, violation of law.” Ibid.
See also id., at 68; Givens v. Zerbst, 255 U. S. 11, 20 (1921); Ver Mehren v. Sirmyer, 36 F. 2d 876, 879-880 (CA8 1929). Just as the silence of the accused in McClaughry could not confer jurisdiction on a court-martial of the Regular Army that was acting in excess of its statutory authority, so here the failure of Gosa to raise his jurisdictional objection before the court-martial could not have conferred upon that tribunal authority that constitutionally could not be conferred. Consequently, his *709failure to object to the jurisdiction of the court-martial that tried him cannot be deemed fatal in this Court.14
B
Moreover, even if O’Callahan were to be treated as merely a procedural rather than as a true jurisdictional decision, application of the doctrine of res judicata would nonetheless be entirely inappropriate in the context of petitioner Gosa’s case since that action was brought by way of a petition for federal habeas corpus. Specifically, I must vigorously disagree with the suggestion, necessarily inherent in Mr. Justice Douglas’ opinion, that the doctrine of res judicata may have some place in the law of federal habeas corpus. In the past, this Court has indicated quite explicitly to the contrary:
“At common law the doctrine of res judicata did not extend to a decision on habeas corpus refusing to discharge the prisoner. The state courts generally have accepted that rule where not modified by statute . . . ; and this Court has conformed to it and thereby sanctioned it ... . We regard the rule as well established in this jurisdiction.” Salinger v. Loisel, 265 U. S. 224, 230 (1924).
See Fay v. Noia, 372 U. S. 391, 423 (1963); Darr v. Burford, 339 U. S. 200, 214 (1950). Indeed, the rule was still *710“well established in this jurisdiction” just a few months ago.15 See Neil v. Biggers, 409 U. S. 188, 190-191 (1972). The federal courts, to be sure, are not without means for *711dealing with repetitious applications for habeas corpus, see, e. g., Salinger v. Loisel, supra, at 231-232; 28 U. S. C. §§ 2244 (a), (b), or with applications raising questions previously litigated in this Court, see 28 U. S. C. § 2244 (c). But no such problems are presented here. Rather, a procedural problem arises in this case because petitioner Gosa failed to assert the “jurisdictional” defect, which he now raises, in seeking leave for a direct appeal to the Court of Military Appeals. This reflects, in my view, a failure on the part of Gosa to satisfy the exhaustion requirement, which is applied in the context of collateral attack on federal habeas corpus, thereby raising a substantial question whether he has waived his right to challenge the “jurisdiction” of the court-martial on habeas corpus.
The exhaustion doctrine evolved in the context of collateral attack on state criminal proceedings. See, e. g., Ex parte Hawk, 321 U. S. 114 (1944); Ex parte Royall, 117 U. S. 241 (1886). It generally requires state petitioners to utilize available state court remedies be*712fore resorting to federal habeas corpus,16 and thus serves both to ensure the orderly functioning of state judicial processes, without disruptive federal court intervention, and to allow state courts to fulfill their roles as co-equal partners with the federal courts in the enforcement of federal law, thus often eliminating the need for federal court action, and avoiding unnecessary friction between state and federal courts. These same considerations inhere in the context of collateral attack in federal court upon the judgments of military tribunals, which constitute a judicial system — a system with its own peculiar purposes and legal traditions — distinct from the federal judicial system much like the independent state judicial systems. Accordingly, this Court normally has required that military petitioners exhaust all available remedies within the military justice system. See Noyd v. Bond, 395 U. S. 683, 693 (1969); Gusik v. Schilder, 340 U. S. 128, 131-132 (1950).17 At the time petitioner Gosa initiated this collateral attack he indeed had not exhausted a military remedy which was formerly available to him *713with respect to the claim he now asserts. But that certainly ought not to be the end of the inquiry.
In Fay v. Noia, 372 U. S. 391 (1963), the Court rejected the position that a state prisoner who had not pursued his state appellate remedies was barred from seeking federal habeas corpus because of his failure to exhaust, where the state appellate remedies were no longer available. The Court concluded, instead, that the exhaustion “requirement refers only to a failure to exhaust state remedies still open to the applicant at the time he files his application for habeas corpus in the federal court.” Id., at 399. The Court established that where there has been a failure to resort to a state court remedy and that remedy is no longer available, the availability of federal habeas corpus would turn on whether there was a deliberate bypass of the state process. Id., at 438. In determining whether such a bypass has occurred, the Court said that “[t]he classic definition of waiver enunciated in Johnson v. Zerbst, 304 U. S. 458, 464 — 'an intentional relinquishment or abandonment of a known right or privilege’ — furnishes the controlling standard.” 372 U. S., at 439.
This Court has never considered the applicability of the nondeliberate-bypass rule in the context of military petitioners. Fay does not speak specifically with respect to such petitioners. Nonetheless, the considerations which argue in favor of tempering the exhaustion requirement with a rule of nondeliberate bypass in the context of state petitioners are equally applicable in the context of military petitioners. Certainly, military petitioners should be encouraged to raise their constitutional claims before available military tribunals in order to ensure the orderly functioning of the system of military justice, to avoid needless federal court action, and to allow military tribunals an initial opportunity to correct *714their own errors. These interests are not subverted, however, by allowing a military petitioner to seek federal habeas corpus on the basis of a claim which he failed to raise before the military courts because he either was unaware of or did not otherwise willingly fail to raise that claim. As with state petitioners, the integrity of the exhaustion requirement is adequately protected by a rule prohibiting a deliberate bypass of an available military tribunal. A more stringent rule would serve only to bar presentation of valid federal claims without any countervailing justification for doing so.
On the facts of this case, I find it impossible to conclude that petitioner Gosa has waived his right to challenge the “jurisdiction” of the court-martial which convicted him of rape on the ground that the offense was not service connected. A valid waiver requires the “intentional relinquishment ... of a known right.” 18 At *715the time of petitioner’s 1967 application for review by the Court of Military Appeals the substantial “jurisdictional” issue that he now raises had yet to be addressed by this Court. While O’Callahan is, to be sure, properly viewed as one further step in the ongoing process of establishing the limits of court-martial jurisdiction, see supra, at 705-706,1 do not think it follows that we should impose a rule of waiver so strict that it requires an individual petitioner to anticipate, at the time he appeals, a particular constitutional ruling of this Court that has yet to be rendered, especially not when the protection of a number of guarantees of the Bill of Rights is at stake. Moreover, where a new constitutional rule has been established following completion of regular proceedings in the military courts, the interests served by the exhaustion requirement can be fully satisfied by requiring that the subsequently identified claim first be presented to the military courts if a means, such as post-conviction relief,19 exists for doing so. Cf. Blair v. California, 340 F. 2d 741 (CA9 1965); Pennsylvania ex rel. Raymond v. Rundle, 339 F. 2d 598 (CA3 1964). Yet if it is clear *716that those courts would reject the claim, such post-conviction resort to the military courts would, of course, be futile and is therefore unnecessary, see Gusik v. Schilder, 340 U. S., at 132-133. This is now the case here, for during the pendency of this action the Court of Military Appeals, in Mercer v. Dillon, 19 U. S. C. M. A. 264, 41 C. M. R. 264 (1970), held that the “jurisdictional” principle announced in O’Callahan did not apply to cases decided before the date of the O’Callahan decision. It therefore became clear that it would be pointless to dismiss petitioner Gosa’s application in order to allow him to present his claim to the military courts,20 and consequently, his challenge to the “jurisdiction” of the court-martial that tried him is now properly before this Court.
Since I then cannot agree with the opinion of either the plurality or Mr. Justice Douglas, I dissent.
Mr. Justice Stewart joins this opinion only as it applies to No. 71-6314. See ante, this page.
See generally Restatement of Judgments § 7, comments at 41-46 (1942).
See 395 U. S., at 265, 267, 269, 272.
In Relford v. Commandant, 401 U. S. 355, 356 (1971), Mr. Justice Blackmun, speaking for the Court, described the O’Callahan decision as follows:
“In O’Callahan ... , by a five-to-three vote, the Court held that a court-martial may not try a member of our armed forces charged with attempted rape of a civilian, with housebreaking, and with assault with intent to rape, when the alleged offenses were committed off-post on American territory, when the soldier was on leave, and when the charges could have been prosecuted in a civilian court.”
A serviceman convicted by a court-martial does, of course, ultimately have access to the federal judicial system by way of a petition for federal habeas corpus. See, e. g., Burns v. Wilson, 346 U. S. 137 (1953); Gusik v. Schilder, 340 U. S. 128 (1950).
Indeed, even if the military voluntarily elected to provide servicemen on trial before courts-martial with the full panoply of procedural rights constitutionally required in civil forums, that would not affect the decision in O’Callahan. Implicit in O’Callahan is the fact that the military system of justice has never been understood to be constitutionally compelled to provide many of the procedural rights afforded by the civilian courts, and thus it would always remain free to provide only that which is constitutionally necessary. It was with an understanding of what is constitutionally required, not of what the military might elect to provide, that the scope of Congress' power under Art. I, § 8, cl. 14, had to be, and was, defined in O’Callahan, see 395 U. S., at 261-262. It is this fact that perhaps best demonstrates the true jurisdictional — as opposed to procedural— nature of that decision.
See also Developments in the Law — Federal Habeas Corpus, 83 Harv. L. Rev. 1038, 1209 (1970). The Court moved beyond the jurisdictional limitation on collateral attacks upon court-martial convictions in Burns v. Wilson, 346 U. S. 137 (1953). See Developments in the Law — Federal Habeas Corpus, supra, at 1215-1216.
Cf. Restatement of Judgments § 7, comment b, pp. 42-43 (1942) :
“There are many situations in which a court lacks competency to render a judgment. Thus, although a State has jurisdiction to grant a divorce of parties domiciled within the State, a decree of divorce rendered by a court which is not empowered to entertain suits for divorce is void. Similarly, a judgment rendered by a justice of the peace is void if under the law of the State such justices are not empowered to deal with the subject matter of the action; as, for example, where the action is one for tort and justices of the peace are given no power except in actions of contract. So also, where a court is given power to deal with actions involving no more than a designated amount, the statute limiting the amount is ordinarily construed not merely to make erroneous a judgment rendered by such a court in excess of its power, but to make such judgment void.”
In Robinson itself, the Cqurt concluded that, in all events, there was no substantial element of reliance since “Waller cannot be said to have marked a departure from past decisions of this Court.” 409 U. S., at 510.
Kinsella v. Singleton, 361 U. S. 234 (1960), Reid v. Covert, 354 U. S. 1 (1957), and Ex parte Milligan, 4 Wall. 2 (1866), dealt with the exercise of military jurisdiction to try civilians, not servicemen. In each case, the Court held that the military lacked jurisdiction to try the civilians.
In Grafton v. United States, 206 U. S. 333 (1907), the Court held that a soldier who had been acquitted by a properly convened court-martial of a charge of homicide growing out of the shooting of a civilian while he was on guard duty in the Phillipine Islands could not thereafter be tried and convicted for the same offense by a civilian court of that Territory. Johnson v. Sayre, 158 U. S. 109 *705(1895), involved the court-martial conviction of a navy paymaster, whom the Court found to be in the naval service of the United States, for embezzling naval funds while serving on a receiving ship of the United States Navy. And in Smith v. Whitney, 116 U. S. 167 (1886), the Court was asked to order that a writ of prohibition be issued against a court-martial convened to try a naval pay inspector essentially for making various contracts not in the best interest of the Navy, for failing properly to enforce contractual agreements with the Navy, for compelling payment of illegal contractual claims against the Navy, and for failing to perform his duties and responsibilities. There can be little question that each of the offenses in Grafton, Johnson, and Smith, was “service connected” within the meaning of O’Callahan. Contrast Relford v. Commandant, 401 U. S., at 365.
Finally, Coleman v. Tennessee, 97 U. S. 509 (1879), involved the court-martial conviction of a soldier for the murder of a civilian woman. The particular circumstances of the murder are not apparent from the Court’s opinion, but it is clear that the crime occurred during the Civil War, that is, during wartime, rather than during peacetime, see id., at 516-517. O’Callahan did not clearly speak with respect to constitutional limits of court-martial jurisdiction during wartime since the offense at issue there had occurred in peacetime, and the plurality does not reach the issue of wartime offenses today, although it arguably is presented in No. 71-1398, see ante, at 685 n. 8.
With regard to the question of official reliance, it has been pointed out that as long ago as 1955 the Departments of Justice and Defense reached an agreement that at least federal offenses committed by servicemen off-post would fall within the jurisdiction of the Justice Department while those committed on-post would be within the jurisdiction of the Defense Department:
“The Departments of Justice and Defense have found it desirable to establish ground rules for determining the forum for trying a serviceman charged with a civil offense in violation of both military and federal law. In general, these rules, which were established by agreement between the Departments in 1955, give to the military department concerned the responsibility of investigating and prosecuting offenses committed by persons subject to the Uniform Code of Military Justice and involving as victims only those persons or their civilian dependents residing on the military installation in question.” Duke & Vogel, The Constitution and the Standing Army: Another Problem of Court-Martial Jurisdiction, 13 Vand. L. Rev. 435, 455 (196.0), citing Army Reg. 22-160, Oct. 7, 1955, implementing Memorandum of Understanding Between the Departments of Justice and Defense Relating to the Prosecution of Crimes Over Which the Two Departments have Concurrent Jurisdiction (July 19, 1955).
Since the plurality opinion does not find it necessary to reach the Secretary’s additional argument in No. 71-1398 that the auto *707theft there at issue was service connected because the offense took place while respondent was absent without leave during wartime, I think it inappropriate for me to express any view on that additional argument at this time.
See also Restatement of Judgments § 7, comment d, p. 45 (1942).
Contrast n. 15, infra.
Mr. Justice Douglas would seem inclined to limit unwaivable jurisdictional flaws to instances in which an accused is “tried by a kangaroo court or by eager vigilantes . . . ante, at 689-690. But the presence or absence of adjudicatory power does not turn only on the fairness of the proceeding afforded by a particular forum; rather, as McClaughry adequately illustrates, jurisdictional competency in the context of courts of limited jurisdiction such as courts-martial necessarily involves the limits of the statutory and constitutional authority that provides the legal underpinnings for such tribunals. See also Hiatt v. Brown, 339 U. S. 103, 111 (1950); and n. 7, supra.
For this reason, I believe that Mr. Justice Douglas’ reliance on Chicot County Drainage District v. Baxter State Bank, 308 U. S. 371 (1940), is clearly misplaced insofar as petitioner Gosa’s case is concerned. Chicot County involved a question concerning the extent of indebtedness on certain municipal bonds whigh had previously been the subject of a federal proceeding to readjust indebtedness under the bankruptcy laws. Following the readjustment proceeding, this Court declared unconstitutional the statute under which the proceeding had been brought, see Ashton v. Cameron County District, 298 U. S. 513 (1936). In Chicot County, this Court then held that the original decree was not open to collateral attack as void by the nonconsenting bondholders who had had notice of the original readjustment proceeding but had there lodged no objection to the court’s jurisdiction.
The decision can be seen as resting simply on the doctrine of res judicata to which the Court referred at points in its opinion, see Chicot County, supra, at 374-375. The plaintiffs in the second action had had a full and fair opportunity to litigate the issue of jurisdiction in the first proceeding, but had failed to do so. At the same time, there had been substantial action taken in reliance on the readjustment plan approved in the first proceeding. New bonds had been sold to the Reconstruction Finance Corporation which had then purchased old bonds in exchange for them. Under these circumstances it was both fair and proper to bar litigation of the jurisdiction issue in the collateral proceeding. Cf. Restatement of Judgments §10 and comment (1942).
But, as has been pointed out, the doctrine of res judicata has no place in federal habeas corpus; rigid rules restricting what questions are open to litigation on collateral attack are inappropriate in the context of judgments affecting personal liberty. There are, of course, legitimate concerns with finality in criminal proceedings — both civilian and military — and with the orderly functioning of independent judicial systems. But we have rules concerning exhaustion, waiver, and non-repetitious application to protect those concerns in the context of federal habeas corpus.
More generally, Chicot County is probably most appropriately interpreted as an early decision concerning the nonretroactive application of a particular decision, nameiy, Ashton. Despite the Court’s *711resort at places to the rubric of res judicata, the presence of substantial reliance on pre-existing law clearly was an important consideration in the Court’s decision not to allow the intervening decision in Ashton to be used to collaterally attack the original plan of readjustment. Furthermore, Chicot County was heavily relied upon by this Court when it gave the principles governing the retroactivity of new procedural constitutional rules full expression in Linkletter v. Walker, 381 U. S. 618, 625-626 (1965); and the case has been cited as a retroactivity decision on a number of occasions since Linkletter, see Chevron Oil Co. v. Huson, 404 U. S. 97, 106 (1971); United States v. U. S. Coin & Currency, 401 U. S. 715, 742-743 (1971) (White, J., dissenting); cf. United States v. Estate of Donnelly, 397 U. S. 286, 293-294 (1970); id., at 299-300 (Douglas, J., dissenting). Viewed then as a precursor of the present-day retroactivity doctrine, Chicot County has no relevance for the threshold question whether Gosa is barred from raising his jurisdictional challenge on habeas corpus because he failed to present it in applying for leave to appeal to the Court of Military Appeals.
This rule does not, however, entitle the state courts to more than one opportunity to consider the same claim. Thus, in Brown v. Allen, 344 U. S. 443, 447 (1953), where the petitioners had presented their federal claims to the state courts on direct review, the Court said, “It is not necessary in such circumstances for the prisoner to ask the state for collateral relief, based on the same evidence and issues already decided by direct review . . . .” Indeed, if the exhaustion requirement were not restricted to providing all levels of the state courts with an opportunity to hear his federal claim, it would effectively bar state prisoners from ever reaching a federal forum in States in which an unlimited number of identical applications for state post-conviction relief are permitted. The exhaustion requirement does not demand such “repetitious applications to state courts.” Id., at 448-449, n. 3.
But see McElroy v. Guagliardo, 361 U. S. 281 (1960); Reid v. Covert, 354 U. S. 1 (1957); Toth v. Quarles, 350 U. S. 11 (1955); Noyd v. Bond, 395 U. S. 683, 696 n. 8 (1969).
Nothing in this Court’s recent decisions in Tollett v. Henderson, 411 U. S. 258 (1973), and Davis v. United States, 411 U. S. 233 (1973), suggests that a different standard should be applied in the context of this case. Tollett involved a collateral attack upon the validity of a guilty plea in light of racial discrimination in the composition of the state grand jury that had indicted Henderson, an objection that had not been raised at the time of the entrance of the plea. Because it was clear that neither Henderson nor his counsel was aware of the claim of discrimination at the time of the plea, the Court agreed that there had been no valid waiver of the claim in traditional terms, see 411 U. S., at 266, but the Court did not consider that determination dispositive in the peculiar context of a collateral attack upon a guilty plea. Rather, the Court ruled that “[t]he focus of federal habeas inquiry is the nature of the advice and the voluntariness of the plea, not the existence as such of an antecedent constitutional infirmity,” ibid. We, of course, do not deal here with the special problem of a collateral attack upon a guilty plea.
In Davis, the Court held that, for purposes of collateral attack, a petitioner had waived his objection to the composition of the grand jury that tried him because he had failed to raise the objection *715before trial as Fed. Rule Crim. Proe. 12 (b) (2) expressly requires. Rule 12 (b)(2) specifies that “[djefenses and objections based on defects in the institution of the prosecution or in the indictment . . . may be raised only by motion before trial” and that failure to do so “constitutes a waiver thereof.” Confronted with a situation in which a specific rule provided “for the waiver of a particular kind of constitutional claim if it be not timely asserted,” 411 U. S., at 239-240, the Court concluded that preservation of the integrity of the Rule demanded that its standard should govern in the context of a collateral attack upon an indictment. This case, however, involves no such “express waiver provision,” id., at 239, and consequently the general waiver principles established by this Court’s previous decisions must control.
See Developments in the Law — Federal Habeas Corpus, 83 Harv. L. Rev. 1038, 1234 (1970); cf. Noyd v. Bond, 395 U. S., at 695 n. 7.
In any case, while his application for habeas corpus was pending in the District Court, petitioner Gosa filed a motion to vacate his conviction and sentence, on the basis of O’Callahan, in the Court of Military Appeals. Subsequent to the denial of relief in the District Court, the Court of Military Appeals, treating petitioner’s motion as a petition for reconsideration, also denied relief. It did so, not on the basis that Gosa had waived the “jurisdictional” question by failing to present it on direct appeal, but on the basis of its previous decision in Mercer holding O’Callahan to be nonretroactive. 19 U. S. C. M. A. 327, 41 C. M. R. 327 (1970). Thus, in all events, it seems clear that Gosa has now adequately exhausted his military remedies and his previous bypass can no longer be deemed a waiver of the “jurisdictional” question, see Warden v. Hayden, 387 U. S 294, 297 n. 3 (1967).