dissenting.
The Court today holds that this case is not moot despite the voluntary dismissal of the prosecution by the Government. It also holds that police on a roving, random patrol may stop and board any vessel, at any time, on any navigable waters accessible to the open sea, with no probable cause or reasonable suspicion to believe that there has been a crime or a border crossing, and without any limits whatever on their discretion to impose this invasion of privacy. Because I cannot agree with either holding, I dissent.
*594rH
It is long settled that a party may not seek appellate review when it has itself sought and obtained entry of a judgment against it, unless it does so solely as a device by which to obtain immediate appellate review of an interlocutory order. E. g., United States v. Procter & Gamble Co., 356 U. S. 677, 680-681 (1958); United States v. Babbitt, 104 U. S. 767 (1882); Evans v. Phillips, 4 Wheat. 73 (1819).
Yet that is precisely what the Court permits the Government to do in this case.1 Respondents were convicted of drug violations and sentenced to prison. The Court of Appeals reversed the judgment on August 3, 1981, holding that the convictions rested on illegally obtained evidence. Rehearing was denied on October 19, and the mandate issued on October 29. On November 20, the Court of Appeals granted the Government’s motion to recall the mandate and stay its reissuance until December 7, pending a petition for writ of certiorari in this Court. The Government, however, permitted that stay to expire without filing the petition, and the *595mandate issued on December 8. On December 21, the Government moved voluntarily in the District Court for dismissal of the indictment under Federal Rule of Criminal Procedure 48(a), and the motion was granted the same day. Not until January 18, 1982, did the Government file its petition for certiorari in this Court.2
Rule 48(a) provides that the Government “may by leave of court file a dismissal of an indictment, information or complaint and the prosecution shall thereupon terminate” (emphasis added). No one has ever challenged the effectiveness of the District Court’s order of dismissal, or sought to set it aside, either by a request for rehearing in that court or by direct review on appeal. Yet the Government, having itself permanently terminated this prosecution, now asks this Court to reinstate respondents’ convictions — convictions for which there is no pending indictment and no extant criminal action. Neither the Government nor the Court provides any adequate explanation of how this is possible.
The Court relies primarily on cases holding that issuance of the mandate of a court of appeals does not necessarily moot a case. Ante, at 581-582, n. 2. That is ordinarily true enough, but it is quite beside the point. The act that terminated this case was not the issuance of the mandate (or the Government’s failure to seek a further stay), but the dismissal of the indictment at the Government’s request. The Court cites Mancusi v. Stubbs, 408 U. S. 204, 205-207 (1972), as support for the proposition that the Court may reinstate respondents’ convictions despite the dismissal. Presumably the Court refers to our holding in Mancusi that “[p]etitioner’s obedience to the mandate of the Court of Appeals and the judgment of the District Court does not moot this case.” Id., at 206 (footnote omitted).3 The unspoken but necessary step in the *596Court’s logic is the Government’s assertion that "the indictment in this case was dismissed solely in order to comply with the court of appeals’ mandate.” Supplemental Brief for United States 3. That assertion, however, is patently false. Not one syllable of the Court of Appeals’ mandate or opinion purported to require the District Court to dismiss the indictment, or to require the Government to move for dismissal. The Court of Appeals held only that respondents’ convictions were infirm because based on inadmissible evidence; it remained open for the Government to retry them on proper evidence, or to seek further review in this Court. The Government points out that it had no other sufficient evidence, and hence as a practical matter it could not have retried respondents. In that circumstance a dismissal of the indictment was indeed a sensible response to the Court of Appeals’ decision, if the Government did not intend to proceed further in seeking to impose criminal liability on respondents. But if, on the contrary, the Government intended to seek a reversal in this Court of the Court of Appeals’ judgment, then there was no reason why it would or should terminate the prosecution by moving under Rule 48(a) for dismissal. Instead, it could, should, and would have proceeded in this Court, allowing the indictment to stand pending our disposition. Neither the *597Government nor the Court draws my attention to anything that would have foreclosed this course of action.4 Plainly, the Government’s motion was based on a decision (presumably later changed) to let the case drop, contenting itself with deportation.
The Court points out that preliminary steps in a prosecution are merged into a conviction and sentence. Ante, at 581-582, n. 2. Again, this is true enough as a general rule, but it is hard to see how it provides any support for the Court’s position. The rule means simply that interlocutory steps are subject to attack on appeal from the final judgment; it has never been meant or taken to undermine the fundamental principle that an indictment is the necessary foundation of and predicate for a felony prosecution, conviction, or sentence. On the contrary, it means just the opposite — that the indictment can be attacked on appeal from the conviction, and if it is defective, the entire conviction and sentence falls. Likewise, if the indictment is dismissed, everything that has been “merged” with it is necessarily included in the dismissal. Where there is no valid indictment pending, “[i]t is of no avail... to say that the court still has jurisdiction of the person and of the crime; for, though it has possession of the person, and would have jurisdiction of the crime, if it were properly presented by indictment, the jurisdiction of the offence is gone, and the court has no right to proceed any further in the progress of the case for want of an indictment.” Ex parte Bain, 121 U. S. 1, 13 (1887).5 Rule 48(a) is but a *598recognition of this principle: Once the indictment is dismissed, “the prosecution shall thereupon terminate.” This prosecution has terminated, and this Court is entirely without power to revive it, or the convictions or sentences that arose out of it and died with it. Hence, because there is no nonadvisory relief that we may grant to the Government, the case should be vacated and remanded with instructions to dismiss as moot.
II
Today, for the first time in the nearly 200-year history of the Fourth Amendment, the Court approves a completely random seizure and detention of persons and an entry onto private, noncommercial premises by police officers, without any limitations whatever on the officers’ discretion or any safeguards against abuse. The Court makes no pretense that its issuance of this maritime writ of assistance is supported by any precedent approving such extraordinary and unregulated powers.6 Instead, it correctly recognizes that *599the relevant precedents are those governing searches or stops of vehicles by police on random patrol or at fixed checkpoints. Almeida-Sanchez v. United States, 413 U. S. 266 (1973); United States v. Brignoni-Ponce, 422 U. S. 873 (1975); United States v. Ortiz, 422 U. S. 891 (1975); United States v. Martinez-Fuerte, 428 U. S. 543 (1976); Delaware v. Prouse, 440 U. S. 648 (1979). But those precedents cannot be read to support or permit today’s holding, for not one of them holds or even hints that a police officer on roving patrol may stop, seize, enter, or search any vehicle, vessel, or person at the whim of the officer. Instead, the cases uniformly hold that any stop or search requires probable cause, reasonable suspicion, or another discretion-limiting feature such as the use of fixed checkpoints instead of roving patrols. If we *600are to reach the merits, therefore, our precedents compel an affirmance.
The Court freely admits that the limitations we have imposed on police discretion were necessary to our holdings in the vehicle-stop cases, ante, at 588, and that the seizure and boarding at issue in this case cannot pass muster under those precedents, ibid. Yet it upholds this seizure, concluding that there are differences between boats and cars sufficient to justify such a blatant departure from solid and recent constitutional precedent.7 There are three basic flaws in the *601Court’s reasoning. First, the Court’s exclusive focus on available tools of investigation puts the cart before the horse; it completely overlooks the primary and overarching concern that has guided our previous decisions — our unqualified and consistent rejection of any “standardless and unconstrained discretion,” Prouse, supra, at 661, that would subject our liberties to the whim of an individual police officer in the field. Second, the supposed factual differences are either insubstantial or of the Government’s own making. And third, it is a non sequitur to reason that because the police in a given situation claim to need more intrusive and arbitrary enforcement tools than the Fourth Amendment has been held to permit, we may therefore dispense with the Fourth Amendment’s protections.
A
In Almeida-Sanchez, we held that police officers on a roving patrol must have probable cause to suspect that a vehicle contains illegal aliens or contraband before they may search it. In Ortiz, we held that the same rule governs searches of vehicles at fixed checkpoints. In either case, the severity of the intrusion and the selective discretion necessarily exercised by police in the field require that that discretion be limited by a requirement of probable cause:
“This degree of discretion to search private automobiles is not consistent with the Fourth Amendment. A search, even of an automobile, is a substantial invasion of privacy. To protect that privacy from official arbitrariness, the Court has always regarded probable cause as the minimum requirement for a lawful search.” Ortiz, supra, at 896 (footnote omitted).
*602In Brignoni-Ponce and Martinez-Fuerte, the Court addressed the limits on police officers’ power to stop vehicles and question the occupants, without searching either vehicles or occupants. These cases were not governed by the probable-cause requirement of Almeida-Sanchez and Ortiz because the police procedures in question were considerably less intrusive than full vehicle searches. Nevertheless, we continued to insist, as we have always done, that there must be some meaningful check on the arbitrary discretion of the police.
In Brignoni-Ponce, the stop in question was made by Border Patrol officers on a roving patrol. We held that such stops are permitted only if the police have a reasonable suspicion that the vehicle contains illegal aliens. As in the vehicle-search cases, we rested primarily on the Fourth Amendment’s command that police discretion be limited by independent constitutional constraints:
“We are unwilling to let the Border Patrol dispense entirely with the requirement that officers must have a reasonable suspicion to justify roving-patrol stops. [T]he reasonableness requirement of the Fourth Amendment demands something more than the broad and unlimited discretion sought by the Government. ... To approve roving-patrol stops of all vehicles in the border area, without any suspicion that a particular vehicle is carrying illegal immigrants, would subject the residents of these and other areas to potentially unlimited interference with their use of the highways, solely at the discretion of Border Patrol officers. [I]f we approved the Government’s position in this case, Border Patrol officers could stop motorists at random for questioning, day or night, anywhere within 100 air miles of the 2,000-mile border, on a city street, a busy highway, or a desert road, without any reason to suspect that they have violated any law.” 422 U. S., at 882-883 (footnote omitted).
*603In Martinez-Fuerte, we held that Border Patrol officers may stop vehicles and question their occupants at fixed checkpoints without probable cause or reasonable suspicion. As the Court recognizes, ante, at 588-589, the reason why reasonable suspicion was required in Brignoni-Ponce but not in Martinez-Fuerte was the additional feature in the latter case that the stops took place at fixed checkpoints rather than on roving patrols. Fixed checkpoints have two major advantages, for Fourth Amendment purposes, over roving patrols: They decrease somewhat the intrusiveness of the stop, and they significantly channel and limit the discretion of the officers and the consequent potential for abuse.
“[WJe view checkpoint stops in a different light because the subjective intrusion — the generating of concern or even fright on the part of lawful travelers — is appreciably less in the case of a checkpoint stop. . . .
“[Cjheckpoint operations both appear to and actually involve less discretionary enforcement activity. The regularized manner in which established checkpoints are operated is visible evidence, reassuring to law-abiding motorists, that the stops are duly authorized and believed to serve the public interest. The location of a fixed checkpoint is not chosen by officers in the field, but by officials responsible for making overall decisions as to the most effective allocation of limited enforcement resources. We may assume that such officials will be unlikely to locate a checkpoint where it bears arbitrarily or oppressively on motorists as a class. And since field officers may stop only those cars passing the checkpoint, there is less room for abusive or harassing stops of individuals than there was in the case of roving-patrol stops.” 428 U. S., at 558-559.
See also Ortiz, 422 U. S., at 894-895.
In Prouse, we reaffirmed our holdings in Brignoni-Ponce and Martinez-Fuerte that stops of vehicles are permissible *604only if made either at fixed checkpoints or on reasonable suspicion. Prouse involved a random, roving-patrol stop of a vehicle for a spot license-and-registration check. As in the prior cases, we relied on the more intrusive nature of random patrols as compared with fixed-checkpoint stops, 440 U. S., at 657, and on the ever-present danger of arbitrariness and abuse posed by the completely discretionary nature of random roving-patrol stops:
“The marginal contribution to roadway safety possibly resulting from a system of spot checks cannot justify subjecting every occupant of every vehicle on the roads to a seizure — limited in magnitude compared to other intrusions but nonetheless constitutionally cognizable— at the unbridled discretion of law enforcement officials. To insist neither upon an appropriate factual basis for suspicion directed at a particular automobile nor upon some other substantial and objective standard or rule to govern the exercise of discretion ‘would invite intrusions upon constitutionally guaranteed rights based on nothing more substantial than inarticulate hunches . . . .’ Terry v. Ohio, 392 U. S. [1,] 22 [(1968)]. When there is not probable cause to believe that a driver is violating any one of the multitude of applicable traffic and equipment regulations — or other articulable basis amounting to reasonable suspicion that the driver is unlicensed or his vehicle unregistered — we cannot conceive of any legitimate basis upon which a patrolman could decide that stopping a particular driver for a spot check would be more productive than stopping any other driver. This kind of standardless and unconstrained discretion is the evil the Court has discerned when in previous cases it has insisted that the discretion of the official in the field be circumscribed, at least to some extent.” Id., at 661 (footnote omitted).
In short, every one of the vehicle-stop precedents on which the Court relies, from Almeida-Sanchez to Prouse, requires *605that a stop or search be supported by either probable cause, reasonable suspicion, or another discretion-limiting feature such as use of fixed checkpoints. But the Court purports to draw from these cases a rule that the police may board any boat, at any time, on any “waters offering ready access to the open sea,” ante, at 588,8 with nothing more to guide them than their unsupported hunch, whim, or even their desire to harass or to flaunt their authority. The boarding at issue here was made by officers on a roving patrol, concededly without any reasonable suspicion of criminal activity. To uphold it is flatly contrary to the square holdings of our cases.
Nor can this departure from Brignoni-Ponce and Prouse be justified by a difference in degree of intrusiveness. The Court asserts that its rule involves “only a modest intrusion,” ante, at 592 (although, the Court admits, not a “minimal” one, ante, at 593). The intrusion is modest, if the comparison is made to a full, detailed search of a vessel and its occupants, which could only be made on probable cause. But the Court’s bland assertion masks the fact that the intrusion at issue here is significantly more severe than those in Brignoni-Ponce and Prouse, which we held permissible only on reasonable suspicion. As in those cases, the stop is made on a roving patrol, so that it cannot claim the more limited intrusiveness of fixed checkpoints. Also as in those cases, there is a large noncriminal maritime traffic that may henceforth be stopped and boarded at random in nearly any waters, at any time, without any reason to suspect that there has been any violation of law. Unlike the earlier cases, however, it does not involve a mere stopping and questioning, cf. infra, at 608, but an actual boarding of a private vessel — more similar to entry of a private house than to the *606stops in Brignoni-Ponce and Prouse. Further, despite the Court’s enthusiasm for identifying differences between boats and cars, it overlooks one obvious difference — the greater expectation of privacy that persons enjoy on boats. A boat, unlike a car, quite often serves as an actual dwelling for its owners, as was apparently true in this case. Even where the owners do not live aboard full-time, a boat may serve essentially the same function as a summer vacation cottage— a residence, albeit a temporary one. In either instance, the occupant would quite reasonably suppose that he was entitled to remain undisturbed by arbitrary government authority. The Court, however, sweeps this expectation aside without a thought.9
Today’s holding thus runs roughshod over the previously well-established principle that the police may not be issued a free commission to invade any private premises without a requirement of probable cause, reasonable suspicion, or some other limit on their discretion or abuse thereof. Here, as in *607Prouse, “[I] cannot conceive of any legitimate basis upon which [a customs officer] could decide that [boarding] a particular [vessel] for a spot check would be more productive than [boarding] any other [vessel]. This kind of standardless and unconstrained discretion is the evil the Court has discerned when in previous cases it has insisted that the discretion of the official in the field be circumscribed, at least to some extent.” 440 U. S., at 661.
B
The Court attempts to justify its departure from Brignoni-Ponce and Prouse by pointing to supposed special law enforcement problems in the maritime setting. I do not accept the premise that such problems permit us to dispense with the Fourth Amendment’s protections against arbitrary police intrusion, see Part II-C, infra. In any event, I am unpersuaded that any sufficiently severe problems have been demonstrated here.
The Court asserts that it is not practicable on water for the police to set up fixed checkpoints such as we approved in Martinez-Fuerte and Prouse. The boarding in this case, however, took place in the Calcasieu Ship Channel, “a separate thoroughfare . . . which all vessels moving between Lake Charles and the open sea of the Gulf must traverse.” Ante, at 582. The Channel bears a strong functional resemblance to the limited-access interstate highways on which the Border Patrol sets up its fixed checkpoints, located so as to funnel most of the relevant traffic through the checkpoints. See Martinez-Fuerte, 428 U. S., at 553. As an opportunity for effective fixed-point inspection, it compares quite favorably to anything likely to 'have been available to the New Castle County, Delaware, patrolman who made the illegal random stop in Prouse. Yet, despite the predictable difficulty of setting up effective checkpoints or even temporary roadblocks in an ordinary urban or suburban network of highways and streets, we held in Prouse that random, roving-*608patrol traffic stops of vehicles are unconstitutional in any setting. There is no justification for departing from that rule in our considerably less extensive system of inland navigable waterways.10
Checkpoints aside, there is no apparent reason why random stops are really necessary for adequate law enforcement. In Prouse, we noted that many, if not all, safety defects are readily detectable by visual means, without any necessity for random stops. 440 U. S., at 660. The same is true of vessels. We also noted that the law enforcement interests at stake could be substantially vindicated by stopping drivers who commit traffic violations. Id., at 659-660. Again, the same is true of vessels. “Smuggling is commonly attended by violation of the navigation laws.” Maul v. United States, 274 U. S. 501, 525 (1927) (Brandéis, J., concurring). Similarly, as we noted in Brignoni-Ponce: “[T]he nature of illegal alien traffic and the characteristics of smuggling operations tend to generate articulable grounds for identifying violators. Consequently, a requirement of reasonable suspicion for stops allows the Government adequate means of guarding the public interest and also protects residents of the border areas from indiscriminate official interference.” 422 U. S., at 883. The case law shows that the same is true of the maritime smuggling trade.11
*609The Court further rests on the fact that vessels, unlike cars, do not carry uniform license plates giving visible evidence of compliance with registration laws. It identifies no reason, however, why that is a necessary or permanent state of affairs. It would be manifestly easy and comparatively inexpensive to provide boats with such means of identification. It is unseemly at best for the Government to refrain from implementing a simple, effective, and unintrusive law enforcement device, and then to argue to this Court that the absence of such a device justifies an unprecedented invasion of constitutionally guaranteed liberties. Moreover, assuming that some check of documents is necessary, the Court does not explain why that need invariably requires the police to board a vessel, rather than to come alongside or to request that someone from the vessel come on board the police vessel. Use of ship-to-shore radio, too, contributes considerably to the Government’s ability to keep track of documentation and registration matters. Cf. Florida v. Royer, 460 U. S. 491, 504-506 (1983) (plurality opinion); id., at 511-512, and n. (Brennan, J., concurring in result).
C
Even if the Court could make a more persuasive showing that there are important differences between vehicles and vessels as to the difficulty of law enforcement, I would not agree with its holding. It simply does not follow that, because the police in particular situations dislike limitations placed on their powers of search and seizure, we may therefore sanction an unprecedented invasion of constitutionally protected liberties.
“The needs of law enforcement stand in constant tension with the Constitution’s protection of the individual *610against certain exercises of official power. It is precisely the predictability of these pressures that counsels a resolute loyalty to constitutional safeguards. It is well to recall the words of Mr. Justice Jackson, soon after his return from the Nuremberg trials:
“‘These [Fourth Amendment rights], I protest, are not mere second-class rights but belong in the catalog of indispensable freedoms. Among deprivations of rights, none is so effective in cowing a population, crushing the spirit of the individual and putting terror in every heart. Uncontrolled search and seizure is one of the first and most effective weapons in the arsenal of every arbitrary government.’ Brinegar v. United States, 338 U. S. 160, 180 [(1949)] (Jackson, J., dissenting).” Almeida-Sanchez, 413 U. S., at 273-274.
Ill
In dissent in Martinez-Fuerte, I expressed my fear that the Court’s decision was part of a “continuing evisceration of Fourth Amendment protections against unreasonable searches and seizures.” 428 U. S., at 567. The majority chided me for my rhetoric and my “unwarranted concern,” pointing out that its holding was expressly and narrowly limited: “Our holding today, approving routine stops for brief questioning ... is confined to permanent checkpoints.” Id., at 566, n. 19. Today the Court breaks that promise. I dissent.
Consider this hypothetical: Peter brings a diversity suit against David, seeking damages for trespass and an injunction against further trespass. The jury awards damages to Peter. On post-trial motions, however, the district judge refuses to enter judgment on the verdict for damages or an injunction; instead, he orders a new trial because he concludes that the verdict rested on improper hearsay evidence. Peter’s lawyer advises him that his chances on retrial are slim; without the supposed hearsay, he has virtually no evidence to support a key element of his case. He advises Peter to pursue an interlocutory appeal under 28 U. S. C. § 1292(a). But Peter decides not to bother further with the case; he files a stipulated dismissal of the complaint under Federal Rule of Civil Procedure 41(a)(1). Thereafter, however, Peter files a notice of appeal, contending that the district judge should have entered judgment on the jury verdict. When the court of appeals asks him about mootness, he asserts that the court should proceed to decide the hearsay issue, because if it holds for Peter it may vacate the dismissal of the complaint and reinstate the jury verdict.
Can there be any doubt that, in this hypothetical case, the court of appeals would throw Peter out on his ear? Yet there is no significant difference between Peter's conduct and that of the Government in this case.
The time for filing was extended by Justice White.
The facts of Mancusi illuminate why that case does not control this one. There, New York had sentenced Stubbs as a second offender, based on an allegedly infirm prior Tennessee conviction. On appeal from a denial *596of federal habeas, the Court of Appeals held that the Tennessee conviction,' and hence the New York sentence, were invalid; accordingly, acting on the Court of Appeals’ mandate, the District Court granted a writ of habeas corpus, ordering that Stubbs be resentenced or released. Before our decision issued, the New York state court complied by resentencing Stubbs. We held that the case was not moot because, if we reversed, the State would be free to reimpose its earlier sentence on Stubbs. (As it happened, the second sentence was the same as the first, but it was still under appeal when our decision was rendered; thus, it was possible that the second sentence would be reversed, leaving the original sentence as the only basis on which New York could impose that punishment.) The key fact in Mancusi was that the State was absolutely required by the District Court’s writ either to resentence Stubbs or to release him; it did not have the option, as the Government did in this case, of simply letting the matter rest pending decision by this Court.
The Government suggests that the Speedy Trial Act, 18 U. S. C. § 3161(e) (1976 ed., Supp. V), somehow foreclosed this. Supplemental Brief for United States 2, n. 1. It is doubtful, however, that a judgment on which certiorari has been granted is “final” within § 3161(d)(2); alternatively, action on the petition for certiorari would likely constitute “other proceedings concerning the defendant” under § 3161(h)(1). In any event, § 3161(e) applies only “[i]f the defendant is to be tried again.” The Government has disclaimed any intention of retrying respondents.
Salinger v. United States, 272 U. S. 542, 549 (1926), hardly limits Bain to its facts, as the Court contends, ante, at 581-582, n. 2; even less does it *598undermine the principle for which I cite the case. Bain held that the Fifth Amendment does not permit amendment of an indictment other than by a grand jury; Salinger held simply that a trial judge may “amend” an indictment by omitting a charge not supported by the evidence at trial. This unsurprising rule is entirely consistent with anything in either Bain or this dissent. It certainly does not in any way contradict Bain’s statement that a live, valid indictment is the sine qua non of any felony prosecution or sentence.
The closest this Court has ever come to granting such unlimited police discretion is in one narrowly limited situation — that of border searches:
“Travellers may be . . . stopped in crossing an international boundary because of national self protection reasonably requiring one entering the country to identify himself as entitled to come in, and his belongings as effects which may be lawfully brought in.” Carroll v. United States, 267 U. S. 132, 154 (1925).
Yet at the same time, we have always stressed the uniqueness of the border-search rule, and have repeatedly pointed out that its rationale cannot acceptably be applied to any other situation:
“It would be intolerable and unreasonable if a prohibition agent were authorized to stop every automobile on the chance of finding liquor and thus *599subject all persons lawfully using the highways to the inconvenience and indignity of such a search. [Tjhose lawfully within the country, entitled to use the public highways, have a right to free passage without interruption or search unless there is known to a competent official authorized to search, probable cause for believing that their vehicles are carrying contraband or illegal merchandise.” Id., at 153-154.
See also, e. g., Almeida-Sanchez v. United States, 413 U. S. 266, 272-274 (1973).
The Government does not contend that the boarding in this case can be justified as a border search. Accordingly, the Court — correctly—does not argue that either the rule or the rationale of the border-search cases has any bearing on this case. In any event, a border search is, in most instances, a fixed-checkpoint stop, sharing the discretion-limiting features of all such stops. See United States v. Ortiz, 422 U. S. 891, 894-895 (1975); United States v. Martinez-Fuerte, 428 U. S. 543, 558-559 (1976); Delaware v. Prouse, 440 U. S. 648, 656-657 (1979); infra, at 603-605. When a border search does not occur at a regular port of entry, it can be made only if it is known that there has in fact been a border crossing. See 3 W. LaFave, Search and Seizure §§ 10.5(d), (e) (1978); cf. United States v. Brignoni-Ponce, 422 U. S. 873, 884 (1975) (Government’s power, if any, freely to stop and question aliens cannot affect Fourth Amendment rights of citizens mistaken for aliens). Hence, the border-search rule does not represent any exception to our uniform insistence under the Fourth Amendment that the police may not be loosed upon the populace with no limits on their ability to stop, seize, or search.
The Court also rests on its assertion that “[i]n a lineal ancestor to the statute at issue here the First Congress clearly authorized the suspicion-less boarding of vessels, reflecting its view that such boardings are not contrary to the Fourth Amendment; this gives the statute before us an impressive historical pedigree.” Ante, at 592; see ante, at 584-587. I cannot agree that every statute enacted by the First Congress must be presumed to be constitutional. See Marsh v. Chambers, 463 U. S. 783, 795 (1983) (Brennan, J., dissenting). Even granting this theory of constitutional adjudication, however, the Court’s historical analysis is self-refuting. The 1790 statute on which it relies, quoted ante, at 584, is by its own terms limited to boardings and searches of ships “if bound to the United States.” 1 Stat. 164 (emphasis added). By contrast, § 48 of the Act, which did authorize customs officers to board and search any vessel without regard to location or entry into the country, was expressly limited to vessels in which customs officers had “reason to suspect any goods, wares, or merchandise subject to duty shall be concealed.” §48, 1 Stat. 170 (emphasis added); cf. Carroll, supra, at 150-151. The Court attempts to explain, away § 48, reasoning that § 48 authorized searches, whereas § 31 authorized only boardings for document checks. Ante, at 585-586, n. 4. Section 31, however, also authorized officers to search an inbound ship, with “free access to the cabin, and every other part of a ship or vessel.” Unless §48 (with its express requirement of reasonable suspicion for searches) is to be read out of the Act, § 31’s broad grant of authority to board and search without suspicion must be read as applying only to ships entering the country — as the language “if bound to the United States” indicates. The section’s further authorization to board and search vessels without suspicion “in any part of the United States” meant merely that customs officials could wait to search a ship until it reached port. In short, § 31 was a border-search statute, applicable only to vessels entering the country. See also n. 6, supra. Thus, as we recognized in Maul v. United *601States, 274 U. S. 501 (1927), it was not until the enactment of the present statute in 1922 that Congress purported to authorize suspicionless boardings of vessels without regard to whether there had been any border crossing. Id., at 505; see id., at 521, 528-529 (Brandeis, J., concurring). Where, then, is the “impressive historical pedigree”?
Since the Court’s holding rests primarily on the need to suppress maritime smuggling, it is necessarily limited geographically to waters accessible to the open sea. The same reasoning requires that today’s rule be limited to such vessels as are capable of having entered the country from the open sea.
The Court points to the system of safety and documentation regulation that vessels must obey. As we pointed out in Prouse, however, the same is true of automobiles, but that does not justify random stops of cars without reasonable suspicion.
“The ‘grave danger’ of abuse of discretion does not disappear simply because the automobile is subject to state regulation resulting in numerous instances of police-citizen contact. ‘[I]f the government intrudes . . . the privacy interest suffers whether the government’s motivation is to investigate violations of criminal laws or breaches of other statutory or regulatory standards.’” 440 U. S., at 662 (citations omitted), quoting Marshall v. Barlow’s, Inc., 436 U. S. 307, 312-313 (1978).
The Court also disparages the significance of the privacy interest in boats by pointing out that, in this case, a private pleasure boat turned out to be engaged in the business of smuggling. Ante, at 592, n. 6. This is precisely the sort of post hoc reasoning, justifying a Fourth Amendment violation by its results, against which we have warned. E. g., Martinez-Fuerte, 428 U. S., at 565. Presumably the Court would not assert that a random, warrantless entry of a private residence on land would be upheld because it turned out that the residence was also being used for some criminal enterprise.
The Court argues that fixed checkpoints are impossible on the open sea. Ante, at 589. Assuming this is true, however, it cannot provide any explanation of why random, suspicionless stops are necessary or permissible on inland waterways such as the Calcasieu Ship Channel. Nor does it explain why, if random stops by roving patrols are necessary, they could not be subjected to some sort of neutral selection system that would decrease the opportunity for arbitrariness or harassment. See Prouse, 440 U. S., at 663-664 (Blackmun, J., concurring).
E. g., United States v. Glen-Archila, 677 F. 2d 809, 813-814 (CA11 1982); United States v. Green, 671 F. 2d 46, 53-54 (CA1 1982); Blair v. United States, 665 F. 2d 500, 505 (CA4 1981); United States v. Streifel, 665 F. 2d 414, 424 (CA2 1981); United States v. D’Antignac, 628 F. 2d 428, 434 (CA5 1980); United States v. Williams, 617 F. 2d 1063, 1077, 1085 (CA5 *6091980); United States v. Zurosky, 614 F. 2d 779, 790 (CA1 1979); United States v. Serrano, 607 F. 2d 1145, 1149 (CA5 1979); United States v. Castro, 596 F. 2d 674, 675-676 (CA5 1979); United States v. Whitmire, 595 F. 2d 1303, 1306 (CA5 1979).