The plaintiffs in this diversity suit seek damages for a nuisance caused by petroleum contamination. Illinois law controls. In 1931 Great Lakes Pipe Line Company, the defendant's predecessor, constructed a six-inch pipeline under a 208-acre tract of farmland in DuPage County. In 1944 the pipeline broke and spilled 30,000 gallons of unleaded gasoline which soaked into the ground and could not be recovered. The land was owned by Thaddeus Milfeld and farmed by his daughter and son-in-law. Great Lakes paid the daughter and son-in-law for the damage caused by the spill, obtaining a release from the son-in-law. The daughter and her siblings later inherited the land and sold it to the plaintiffs' predecessor in title. In 1956 Great Lakes built a second, larger (twelve-inch) petroleum pipeline under the land. Ten years later Great Lakes sold all its pipeline assets to the defendant, expressly retaining all liabilities not listed on its balance sheet. Actual or contingent liabilities arising from damage to land through which the pipelines ran were therefore retained. In 1985 the defendant sold the six-inch pipeline to a fiber optics company and it ceased being used to carry petroleum products.

We come now to the immediate antecedents of the suit. In 1986, the plaintiffs sold an option to a real estate developer to buy the land for the creation of an industrial park. The developer, who eventually exercised the option for some $7 million, made soil borings which revealed petroleum contamination. The plaintiffs retained a contractor named Eiler to clean up the contamination. Eiler in turn retained Testing Service Corporation to take soil samples, and TSC hired NET Midwest to test the samples. After subjecting the samples to chromatographic analysis, NET reported to TSC, which reported to Eiler, that the only petroleum product in the samples was unleaded gasoline. But NET did not give TSC, Eiler, or the plaintiffs the actual chromatograms on which NET's conclusion concerning the nature of the contamination was based.

Satisfied that the contamination had been caused by the unleaded gasoline spilled in 1944, Eiler so represented to the Illinois Environmental Protection Agency, which approved a cleanup plan designed and executed by Eiler at a cost to the plaintiffs of more than $1 million. This suit, filed in July 1989, seeks to recover that cost. Although the complaint alleges petroleum contamination without specifying the type of petroleum product or the date of the break or leak that caused the contamination, pretrial discovery brought to light no indication of any break or leak other than the break that had occurred in 1944. In response to the defendant's discovery request for "all documents referring to, reflecting or relating to any contamination of the Property," the plaintiffs did not hand over the chromatograms, as they had never been in their possession, or Eiler's, or even TSC's.

Discovery, other than of prospective expert witnesses, closed in April 1991. The following month the plaintiffs moved to amend their complaint to add a new claim of contamination, based on the discovery of diesel fuel in two stormwater retention ponds. The judge denied the motion, ruling that it could be made the basis of a new suit if the plaintiffs wanted. The discovery of the diesel fuel had prompted the plaintiffs in March 1991 to arrange for the taking of additional soil samples, which were subjected to chromatographic tests that revealed the presence of diesel fuel. In July, the plaintiffs' expert, Ball, obtained from NET the chromatograms that the laboratory had prepared at TSC's request back in 1989, and according to Ball these chromatograms, too, reveal the presence of diesel fuel in the soil. If Ball is correct, NET had either failed to analyze the chromatograms fully back in 1989 or failed to communicate its analysis fully to TSC.

The plaintiffs submitted copies of the 1989 chromatograms to the defendant forthwith (it had already submitted copies of the 1991 ones to them), only to be met by a motion to exclude the chromatograms from evidence as a sanction for the plaintiffs' having failed to produce them in response to the original document request; for they had been in existence then. The district judge granted the motion--and excluded the 1991 chromatograms to boot. He thought the plaintiffs' failure to have produced the 1989 chromatograms in 1989 was inexcusable; that the plaintiffs were estopped by their representations to the defendant and to the Illinois Environmental Protection Agency to change the theory of their case from a 1944 gasoline spill to a fairly recent (date unknown) spill of diesel fuel; that the new evidence was weak; and that it was too late for the plaintiffs to change the theory of their case. The last must have been the reason the judge excluded the 1991 chromatograms as well--they were pertinent only to a claim of contamination by diesel fuel. With the case now confined to the 1944 spill, the judge granted summary judgment for the defendant on the ground that the defendant bore no liability as the successor to Great Lakes, the owner of the pipeline in 1944.

The easy issue is the correctness of the judgment with respect to the 1944 spill. That was long before the defendant became the owner of the pipeline. The general rule in Illinois as elsewhere is that the purchaser of assets does not acquire the seller's liabilities unless he agrees to do so. Nilsson v. Continental Machine Mfg. Co., 251 Ill.App.3d 415, 190 Ill. Dec. 579, 581, 621 N.E.2d 1032, 1034 (1993); Myers v. Putzmeister, Inc., 232 Ill.App.3d 419, 173 Ill. Dec. 130, 596 N.E.2d 754 (1992). If the liabilities always went with the assets, it would be difficult to sell assets because the purchaser would not know what he was getting. He might be "buying" a lawsuit the expected cost of which exceeded the value of the asset purchased, yet it would be too late for him to back out of the sale or renegotiate the price. Manh Hung Nguyen v. Johnson Machine & Press Corp., 104 Ill.App.3d 1141, 60 Ill. Dec. 866, 872, 433 N.E.2d 1104, 1110 (1982).

The rule permitting assets to be sold separately from liabilities is part of a large family of rules aimed at facilitating transactions by clearing clouds on titles. Another member of the family is the rule that a bona fide purchaser for value takes free of certain claims against the seller in respect to the good sold. See, e.g., UCC Sec. 2-403. If, however, the "sale" is simply a corporate reorganization that leaves real ownership unchanged, the liabilities go with the assets. Nilsson v. Continental Machine Mfg. Co., supra; State ex rel. Donahue v. Perkins & Will Architects, Inc., 90 Ill.App.3d 349, 45 Ill. Dec. 696, 699, 413 N.E.2d 29, 32 (1980); Plaza Express Co. v. Middle States Motor Freight, Inc., 40 Ill.App.2d 117, 189 N.E.2d 382, 385 (1963). And when one corporation is merged into another, the acquiring corporation gets the liabilities of the acquired one along with the assets. Robinson v. KFC National Management Co., 171 Ill.App.3d 867, 121 Ill. Dec. 721, 725, 525 N.E.2d 1028, 1032 (1988); 805 ILCS 5/11.50(a)(5). If it did not, the transaction would be a sale of assets rather than a merger.

We confess to some puzzlement as to why liabilities are retained when the assets sold constitute an entire business, a "going concern." A seller who is exiting from a business doesn't want to be plagued by lawsuits afterward and may not even retain the organizational capacity to defend against them. It is true that a disclaimer of liability is good only against the purchaser, not against a nonconsenting third party. Assignment makes the assignee another obligor; it does not let the assignor off the hook. But it reduces the probability that he will be sued. A further consideration in the case in which the entire business is sold is that the buyer is likely as a matter of ordinary prudence to have investigated the business's history before buying it and through that investigation to have discovered the seller's contingent liabilities, and he is therefore less likely to be surprised by a subsequent lawsuit than the purchaser of a particular asset would be. Of course not all contingent liabilities are foreseeable. The requirements for cleaning up polluted land have become much more stringent since 1966, when Great Lakes sold its pipeline business to the defendant; and it is unlikely that the parties anticipated the political and legal changes that are responsible for that greater stringency.

Still another consideration is that when an entire business is sold the seller may no longer be able to pay a judgment. This is clearest in the case where after the sale of all its assets a corporate seller distributes the proceeds of the sale to the shareholders and dissolves. If the purchaser is not liable, the transaction will have externalized the costs of the seller's acts that gave rise to liability. This consideration may help explain why liabilities are retained rather than transferred when a business operated as a proprietorship rather than in the corporate form is sold, even though all the assets used in the proprietorship are sold. James J. White & Robert S. Summers, Uniform Commercial Code Sec. 19.1, p. 755 (2d ed. 1980). Unlike a corporation that sells its entire business and dissolves, rendering suit against it difficult (to say the least), a proprietor does not dissolve when he sells his business.

The common law rule that the sale of assets does not carry the seller's liabilities with them is qualified by bulk-sale statutes. And it is merely a default rule--a rule that courts use to complete contracts when the parties have failed to specify the allocation of some risk. If, not trusting the default rule, or wanting to vary it, the parties have specified whether liabilities are to be retained by the seller or assumed by the buyer, the court will enforce the specified allocation and the default rule drops out. American Xyrofin, Inc. v. Allis-Chalmers Corp., 230 Ill.App.3d 662, 172 Ill. Dec. 289, 297-98, 595 N.E.2d 650, 658-59 (1992). Neither the default rule nor an express retention of liabilities will be enforced, however, if there is some element of fraud against present or prospective tort or contract or other creditors of the seller. To avoid a judgment in an impending lawsuit or avalanche of suits the seller might have sold all its assets to a new corporation owned by its predecessor's owners and retained all its liabilities in an assetless shell; if so, then as in Wolff v. Shreveport Gas, Electric Light & Power Co., 70 So. 789, 794-95 (La.1916); cf. Plaza Express Co. v. Middle States Motor Freight, Inc., supra, 40 Ill.App.2d 117, 189 N.E.2d at 385, the successor corporation would be liable. Or, again to avert impending financial destruction through litigation, the original enterprise might have sold its assets to a third party, retained the liabilities in the shell, and distributed the proceeds of the sale to its shareholders. The purchaser would be liable, though only up to the value of the assets (since any debt above that value would have been dischargeable in bankruptcy), provided that he knew about the suit. Cf. In re International Paper Co., 961 F.2d 558, 559 (5th Cir.1992).

There are few cases under the fraud exception, partly because the corporate-reorganization rule noted earlier takes care of a Wolff -type corporate shell game and partly because creditors prefer to cast these cases as suits to set aside a fraudulent conveyance. It has been suggested, indeed, that the fraud exception to the nonliability of successors is merely an application of the law of fraudulent conveyances. 15 Fletcher Cyclopedia of the Law of Private Corporations Secs. 7125, 7129 (1990 rev. ed.). We need not pursue the issue. There is no suggestion of fraud here; so the express retention of liability by Great Lakes ends the case so far as the 1944 break is concerned. Even without that express retention the defendant would be off the hook. For while Great Lakes sold its entire pipeline business to the defendant rather than merely selected assets, the default rule as we have said is that liabilities are retained, not transferred, unless the sale is pursuant to a corporate reorganization or merger.

Even less need we consider the argument that as an original matter a seller, whatever the intention behind the transaction (and so regardless of whether there is fraud), ought not be allowed to render itself judgment-proof by selling all its assets and distributing the proceeds of the sale, while retaining all liabilities, thus shielding the buyer; that in such a case the liabilities should follow the assets into the hands of the purchaser, or of the seller's owners, so that the holder of a judgment against the seller will be able to collect on it. This is not the law, with irrelevant exceptions that include special rules for collective bargaining agreements and in some states for products liability suits, and the "trust fund" doctrine, recognized in most states, which imposes continued liability for a brief time on former shareholders of a dissolved corporation for the corporation's debts. Blankenship v. Demmler Mfg. Co., 89 Ill.App.3d 569, 44 Ill. Dec. 787, 789-90, 411 N.E.2d 1153, 1155-56 (1980); In re RegO Co., 623 A.2d 92, 95 (Del.Ch.1992); Mark J. Roe, "Mergers, Acquisitions, and Tort: A Comment on the Problem of Successor Corporation Liability," 70 Va.L.Rev. 1559, 1564 n. 15 (1984). The issue is doubly academic because Great Lakes did not dissolve or otherwise render itself judgment-proof after the sale of its pipeline business.

It is triply academic. The statute of limitations for bringing suit against either Great Lakes or a successor in respect of the 1944 spill expired many years ago. It is true that a statute of limitations does not begin to run in Illinois until the plaintiff (or his predecessor in title) discovers or should have discovered the injury on which he wants to base his suit. Knox College v. Celotex Corp., 88 Ill. 2d 407, 58 Ill. Dec. 725, 728-29, 430 N.E.2d 976, 979-80 (1981); Singletary v. Continental Illinois National Bank & Trust Co., 9 F.3d 1236, 1240 (7th Cir.1993). But it is so highly improbable that Milfeld's daughter and son-in-law failed to inform Milfeld--who lived just half a mile from the farm--of the large oil spill for which Great Lakes was offering compensation that, all witnesses being dead so that only the probabilities of the situation are available to guide decision, no reasonable factfinder could conclude that the statute of limitations had not begun to run when the spill occurred.

The difficult question is whether the plaintiffs should have been allowed in 1991 to expand this lawsuit to take in the possibility that some part of the contamination of their land was due to a spill of diesel fuel. They cannot be criticized for having failed to produce the 1989 chromatograms when the defendant first requested all test results. The request was broad enough to have encompassed any chromatograms that the plaintiffs had; and the plaintiffs do not deny that the request required them to produce every pertinent document in their "possession, custody or control" within the meaning of Fed.R.Civ.P. 34(a)(1), even if it was not in their physical possession. But they did not have custody or control: they could not order NET to surrender the chromatograms to them. Searock v. Stripling, 736 F.2d 650, 653-54 (11th Cir.1984); Gerling Int'l Ins. Co. v. Commissioner, 839 F.2d 131, 139-40 (3d Cir.1988). The plaintiffs disclosed to the defendant the identities of Eiler, TSC, and NET as sources of the information on which the suit was based, and if the defendant had wanted pertinent documents in the custody or control of any of those entities it had only to issue a subpoena duces tecum. Fed.R.Civ.P. 34(c), 45. It did not do so. The plaintiffs could no doubt have asked NET to give it the chromatograms; judging from what happened later, NET would have complied; and maybe if it had balked, the plaintiffs could have bought the chromatograms from it. But the fact that a party could obtain a document if it tried hard enough and maybe if it didn't try hard at all does not mean that the document is in its possession, custody, or control; in fact it means the opposite.

The district judge gave additional grounds, however, for refusing to allow the plaintiffs to press their alternative theory. One was that the evidence for it was very weak. The stormwater retention ponds in which the diesel fuel was first spotted drain a larger area than the plaintiffs' property and anyway would be more likely to collect surface spills than underground ones. A recent spill from the defendant's pipeline (or pipelines, for it had two under the plaintiffs' land until 1985) would, moreover, almost certainly have included some leaded petroleum products; but no lead has ever been found in the plaintiffs' land. And pretrial discovery turned up no record of any post-1944 break, leak, or spill even though the owner of a pipeline has a powerful self-interest, unrelated to liability concerns, in identifying and correcting a leak, since a leak causes a loss of product and resulting loss of revenue to the pipeline company. A leak could go undiscovered for a time, but modern pipelines contain electrical detection systems which make this unlikely. Finally, Eiler's original conclusion that the contamination had been caused by the 1944 spill was based on some 400 soil borings and was concurred in by the public agency responsible for the elimination of such contamination. The inference seems inescapable that the "discovery" of diesel fuel contamination was spurred by a well-founded concern that the plaintiffs' claim based on the 1944 gasoline spill was doomed by Great Lakes' retention of liability and by the expiration of the statute of limitations.

Another ground that the district judge offered for his ruling was judicial estoppel. A litigant is forbidden to obtain a victory on one ground and then repudiate that ground in a different case in order to win a second victory. Davis v. Wakelee, 156 U.S. 680, 689-91, 15 S. Ct. 555, 558-59, 39 L. Ed. 578 (1895); Chrysler Motors Corp. v. International Union, Allied Industrial Workers of America, 2 F.3d 760, 764 (7th Cir.1993); Astor Chauffeured Limousine Co. v. Runnfeldt Investment Corp., 910 F.2d 1540, 1547-49 (7th Cir.1990); 18 Charles Alan Wright, Arthur R. Miller & Edward H. Cooper, Federal Practice and Procedure Sec. 4477 (1981). Though called judicial estoppel, the doctrine has been applied, rightly in our view, to proceedings in which a party to an administrative proceeding obtains a favorable order that he seeks to repudiate in a subsequent judicial proceeding. Smith v. Pinner, 891 F.2d 784, 787 n. 4 (10th Cir.1989) (per curiam); Parisi v. Jenkins, 236 Ill.App.3d 42, 177 Ill. Dec. 496, 503-04, 603 N.E.2d 566, 573-74 (1992); Department of Transportation v. Coe, 112 Ill.App.3d 506, 68 Ill. Dec. 58, 60, 445 N.E.2d 506, 508 (1983); Czajkowski v. City of Chicago, 810 F. Supp. 1428, 1435-36 (N.D.Ill.1992). It was very much in the plaintiffs' interest to persuade the Illinois EPA that the contamination of their land was due to a spill of unleaded gasoline in 1944, because the agency has more stringent requirements for cleaning up land that is contaminated by lead. Having fended off the agency, the plaintiffs want now to reverse course and recover their clean-up costs by persuading a court that what was really cleaned up was a much later diesel-fuel spill.

True, this about-face might not make the defendant on balance worse off than it would have been had the plaintiffs put their money on the diesel-fuel hypothesis from the start. It is not a case of a plaintiff's seeking double recovery by taking inconsistent positions in subsequent suits. All that the plaintiffs are seeking by way of damages is the expense they incurred to clean up the land. If the Illinois EPA had imposed more stringent requirements on the clean up, the plaintiffs would have incurred higher costs and would be seeking higher damages. But the objective of the doctrine of judicial estoppel is not just to protect the party in the suit in which his opponent seeks to repudiate an earlier position successfully asserted. It is also to prevent situations from arising in which one of two related decisions has to be wrong because a party took opposite positions and won both times. If the plaintiffs are right this time, and the contamination that they cleaned up was due to a spill of diesel fuel, then they failed to comply with the state's environmental protection regulations. If they did comply fully with those regulations, it means they are wrong this time and there was no spill of diesel fuel. By making them choose one position irrevocably, the doctrine of judicial estoppel raises the cost of lying. Alternative pleading is permitted, so a party does not have to make a binding election of positions until the first suit goes to judgment. Harbor Ins. Co. v. Continental Bank Corp., 922 F.2d 357, 364 (7th Cir.1990). But he is forbidden to equivocate beyond that point.

This case is further complicated, however, by the possibility that the opposite positions are not really opposite. It may simply be that in 1989, when the plaintiffs persuaded the Illinois EPA that the contamination was due to the old spill, this was the best information available to them and it was not until the diesel fuel showed up in the retention ponds two years later that there was any reason to suspect a spill more recent than 1944. There would be no purchase at all for invoking judicial estoppel if the diesel fuel had been spilled after the Illinois EPA's ruling. But we do not understand the plaintiffs to be arguing that the spill is that recent; the fuel in the ponds may be symptomatic of a spill that occurred earlier, though more recently than 1944.

The plaintiffs should of course have alerted the state agency to the possibility of additional contamination. Their stated reason for not doing so--that none of the additional tests revealed lead--weakens their new theory, because pipelines don't run just diesel fuel and a recent spill would almost certainly have included some leaded products. However that may be, the doctrine of judicial estoppel is not an absolute bar to obtaining legal relief on the basis of new information, even if inconsistent old information had gotten the party an advantage in some other proceeding. It would be odd to apply this rather esoteric though we think salutary doctrine more strictly than collateral estoppel or law of the case, related bars to changing one's position after it has been adopted by a court. At least where there is no issue of double recovery (an important qualification to which we shall return shortly), if the court in the second suit is satisfied that the position adopted in the first suit was clearly wrong yet had been advanced in good faith by the party now sought to be estopped to repudiate it, the court is not required to apply the doctrine of judicial estoppel. In re Cassidy, 892 F.2d 637, 642 (7th Cir.1990).

Whether the court would be permitted to apply the doctrine in such a case if the party sought to be estopped had been fully compensated in the earlier proceeding in which it had taken the opposite position is an interesting question, although one unnecessary to resolve in this case. Judicial estoppel is strong medicine, and this has led courts and commentators to characterize the grounds for its invocation in terms redolent of intentional wrongdoing. Comment, "Precluding Inconsistent Statements: The Doctrine of Judicial Estoppel," 80 Nw.U.L.Rev. 1244, 1264-65 (1986). But barring double recovery hardly seems an excessive sanction for careless as distinct from deliberate reliance in an earlier proceeding on a ground now argued to be mistaken. To base double recovery on the plaintiff's own carelessness would be a strange inversion of the principles of unjust enrichment. But that is not an issue here, and the judge's invocation of the doctrine of judicial estoppel, like his sanctioning the plaintiffs for failing to produce the chromatograms in response to the defendant's discovery request, was unwarranted.

All this said, we think the district judge acted within the bounds of reason in refusing to allow the suit to be expanded to take in the claim of diesel-fuel contamination. It was a weak claim, seemingly a desperate response to the increasing remoteness, legal as well as chronological, of the 1944 spill. And the plaintiffs were seeking to make a fundamental change in the theory of their case after two years of litigation, including extensive pretrial discovery.

Most important, it was not a matter of introducing new evidence, or changing merely the legal conceptualization of their claim, but of injecting an entirely new and separate claim. The original suit--everyone had understood despite the vagueness of the complaint--had been based on damage caused in 1944 by a large spill of unleaded gasoline. The new evidence had nothing to do with the 1944 spill and resulting damage, except that the defendant could be expected to cite that spill as the real cause of any contamination of the plaintiffs' land. Although the date of the hypothetical diesel-fuel spill has never been pinpointed, the plaintiffs describe it as fairly recent and let us suppose, simply to make analysis concrete, that it occurred in 1980. A petroleum spill in 1980 is a different tort from a petroleum spill in 1944. That it came from the same pipeline and polluted the same land no more makes it part of the earlier tort than the fact that A had assaulted B in 1944 and again in 1980 would make the two assaults one tort rather than two. Suppose B had sued A, claiming that an assault (not otherwise specified) by A had caused brain damage to B first discovered in 1985; and suppose the parties had assumed (and B had made representations to the district attorney) that the assault that did the damage was the one that occurred in 1944. Two years into the case B decides that the 1944 assault probably had not caused the damage to his brain--it was the 1980 assault that had done so--and he asks the judge for permission to introduce evidence concerning that assault. It would be entirely within the judge's power to say to B: If you want to bring a case based on the 1980 assault you're free to do so; but the case you have brought is based on the 1944 assault and I will decide that case without allowing it to become encumbered with a separate claim. Gorenstein Enterprises, Inc. v. Quality Care-USA, Inc., 874 F.2d 431, 437-38 (7th Cir.1989); Kier v. Commercial Union Ins. Cos., 808 F.2d 1254, 1258 (7th Cir.1987).

Since the decision whether to permit a complaint to be amended is confided to the district judge's discretion, id., rather than to our own, we cannot be entirely comfortable in affirming, because we cannot be entirely confident that the judge's decision was not affected by misconceptions concerning both the doctrine of judicial estoppel and the culpability of the plaintiffs in failing until 1991 to produce the 1989 chromatograms in response to the defendant's document request. But our confidence is restored by the remarks that the judge made in open court when he denied the motion to amend the complaint. He said that when "you [the plaintiffs] offer something which essentially changes the scope of liability and increases the costs, you are in a very real sense being unfair to your opponent, who had they known about this earlier on may very well have decided that all of this was not worth the candle and this may very well have settled, at less cost to them, at less cost to you and at less time consumed in court, although in all honesty the time consumed by the court in this case has not been particularly great." There is no reference to nonproduction of the 1989 chromatograms or to judicial estoppel. And it undoubtedly is true that a more diligent investigation into the nature and possible causes of the contamination before the plaintiffs filed their complaint would have shortened the litigation, economized on the parties' expenses and the judge's time, and possibly facilitated settlement. For failure to exercise such diligence the sanction imposed by the judge, refusing to allow the complaint to be amended, was mild, since the plaintiffs can bring a new suit in an effort to vindicate their new theory. They can do so because, for the reasons that we have indicated, the doctrine of judicial estoppel would not bar such a suit. The suit would not be barred by res judicata either, since it would arise from a different act, at a different time, causing a different harm, from the first suit--as the defendant itself emphasizes, so that for it to plead res judicata in the next suit would invite application of the doctrine of judicial estoppel against it! And the statute of limitations is not an obstacle. The plaintiffs did not discover the new injury until 1991. They cannot be faulted for NET's failure to discover it or if it did discover it to communicate its discovery to them. So they can obtain all the relief to which they are legally entitled by filing a new suit without the clutter and confusion that turning the old suit 180 degrees in mid-course would invite.

We have thus far been treating the issue as whether the district judge abused his discretion in refusing to permit the complaint to be amended. But that is only part of the issue. He did refuse to permit the complaint to be amended but he also and separately precluded the plaintiffs from using as evidence either set of chromatograms. By doing this he prevented them from developing their new claim, that of a recent spill of diesel fuel. The plaintiffs didn't have to amend their complaint (though they tried to anyway) to add the claim; the complaint was sufficiently vague to encompass it.

We have to compare two situations in order to evaluate the judge's preclusion order. In the first, as is increasingly common despite the "notice pleading" philosophy of the Federal Rules of Civil Procedure and the specific admonition of Rule 8(a) that the statement of the claim in the complaint shall be "short and plain," the complaint is highly specific. So, if the plaintiff in the course of pretrial discovery comes up with a new claim, he will have to get his complaint amended if the pleadings and the proof are to be conforming. In the second situation the statement of "the claim" in the complaint is sufficiently general to encompass a variety of claims, and then no amendment is needed to add a new claim provided it falls within the broad range staked out by the complaint. But the defendant may object to the injection of a new claim after there has been extensive discovery during which that claim was never so much as hinted at. Whatever power a judge has to refuse to amend the complaint in the first situation he should also have in ruling on a request by the defendant in the second situation to exclude the new claim by keeping out the evidence on which it is based. And since the guidelines to the exercise of judicial discretion are more developed in the first situation than in the second, we can follow them here even though this is a case of the second as well as of the first type. Undue delay, undue prejudice to the opposing party, and futility of amendment (because the claim sought to be added has no merit) are all reasons for a district judge's deciding that allowing the complaint to be amended would not serve the interests of justice. Fed.R.Civ.P. 15(a); Foman v. Davis, 371 U.S. 178, 182, 83 S. Ct. 227, 230, 9 L. Ed. 2d 222 (1962). All are present here, justified the denial of the motion to amend the complaint, Johnson v. Methodist Medical Center, 10 F.3d 1300, 1304 (7th Cir.1993); Jones v. Psimos, 882 F.2d 1277, 1285 (7th Cir.1989); Smith v. Duff & Phelps, Inc., 5 F.3d 488, 494 (11th Cir.1993); Berger v. Edgewater Steel Co., 911 F.2d 911, 924 (3d Cir.1990); Anderson v. USAir, Inc., 818 F.2d 49, 57 (D.C.Cir.1987), and therefore equally justified the preclusion order.

It is true that the original theory of the Federal Rules of Civil Procedure was that the plaintiff ought to be permitted to fumble around searching for a meritorious claim within the elastic boundaries of a barebones complaint until the final pretrial conference. No judge or lawyer in this age of crowded dockets takes that completely seriously and in any event it is not a reversible error for the judge to insist that a separate claim be made the subject of a separate suit, especially when the original claim was ripe for judgment.

AFFIRMED.