The plaintiff brought suit seeking a declaration that it had not violated a contract to which it and the defendant, along with a third firm, are parties. Jurisdiction is based on diversity of citizenship, and the governing substantive law is that of the state of Washington, though no peculiarities of Washington law have been drawn to our attention. The defendant counter-claimed, charging that the plaintiff had indeed broken the contract. On cross-motions for summary judgment, the district court ruled for the defendant and entered a permanent injunction against the plaintiff, precipitating this appeal.

The contract involves a medical device known as a stent. The narrowing of an artery, as by atherosclerosis, is called "stenosis" and one way of treating it is by balloon angioplasty, a procedure in which a small balloon is inserted into the affected artery to press against the artery wall restoring it to its normal dimensions. The stent is a metal tube that encloses the balloon and remains in the artery after the procedure. Yet sometimes despite the stent the stenosis reappears — this is called "restenosis" — and there is medical opinion that the likelihood of this happening can be reduced by coating the stent with a suitable drug. One candidate to be such a drug is paclitaxel, which gained fame as an anticancer drug under the trade name Taxol. The patent rights for use of paclitaxel on stents are held by a Canadian company called Angiotech Pharmaceuticals. Angiotech does not manufacture either stents or drugs, and so it decided to license the use of paclitaxel for coating stents. It granted "coexclusive" licenses to Cook Incorporated and Boston Scientific Corporation (BSC), firms involved in the development of drug-coated stents for preventing restenosis. Each license grants the licensee "worldwide right[s] and license to use, manufacture, have manufactured, distribute and sell, and to grant sublicenses to its Affiliates to use, manufacture, have manufactured, distribute and sell [paclitaxel] ... solely for use in [stents]." The licenses are exclusive in the sense that Angiotech promises not to license the use of paclitaxel for coating stents to any other firm, but coexclusive in the sense that each licensee has the same rights. Critically, the licenses forbid the licensee to assign his license, or to grant a sublicense to anyone except an affiliate, unless all parties to the two licenses, which is to say Angiotech, BSC, and Cook, agree. We'll call this provision the "anti-assignment" clause, although it covers sublicenses as well. Why the distinction between assignment to an affiliate, which is forbidden, and sublicensing an affiliate, which is permitted, is unclear, since, while an assignment and a sublicense are not identical, a sublicense can be drafted in such a way as to have the same effect. Finance Investment Co. v. Geberit AG, 165 F.3d 526, 531-32 (7th Cir.1998); Black Clawson Co. v. Kroenert Corp., 245 F.3d 759, 765 (8th Cir.2001); Prima Tek II, L.L.C. v. A-Roo Co., 222 F.3d 1372, 1377 (Fed.Cir. 2000).

The two licenses were granted in a single contract, so that both the licensees and Angiotech are contractually bound to one another. The contract provides for the arbitration of disputes arising under it, but the parties have waived that provision. Angiotech is not a party to this suit; it may be indifferent to the outcome or reluctant to take sides in a dispute between its most important partners in the development of restenosis-resisting stents.

Why coexclusive licenses? No evidence has been presented or arguments made concerning the reason for the coexclusive feature. The lawyers either have not bothered to inquire or have not bothered to inform us or the district judge concerning the commercial setting of such a contract. They seem insensitive to the importance to the sound interpretation of contracts of understanding the business purpose served by a contract's provisions, and to the limitations of generalist judges' knowledge of the customs and practices of specific industries. We are left to speculate, having found no secondary literature on coexclusive licenses either.

A patentee's choice between granting exclusive and nonexclusive licenses is similar to a seller's choice between granting exclusive and nonexclusive rights to his dealers. The dealer who is granted an exclusive right will have an enhanced incentive to devote his sales efforts to the seller's product. Dealers who do not receive exclusive rights will have enhanced incentives to minimize their margins by competing among themselves, thus maximizing the price that the seller can charge. Consumers' demand is a function of the dealer's as well as the original seller's profit margin. For both margins are components of the retail price, and so the lower the dealer's margin is, the lower that price will be, the more therefore will be sold, and so the greater will be the original seller's total profits.

Thus a patentee can ordinarily be expected either to grant nonexclusive licenses in order to exploit the effect of competition in minimizing the licensees' margins or to grant an exclusive license in order to encourage the licensee to invest in the further development of the licensed process or product by protecting the licensee from the competition of other licensees, which might prevent the licensee from recouping his investment. John W. Schlicher, Licensing Intellectual Property: Legal, Business, and Market Dynamics 69-71 (1996). There are other considerations bearing on the choice between exclusive and nonexclusive licensing as well, see, e.g., Michael L. Katz & Carl Shapiro, "How to License Intangible Property," 101 Q.J. Econ. 567 (1986); Carl Shapiro, "Patent Licensing and R&D Rivalry," Am. Econ. Rev. Papers & Proceedings, May 1985, pp. 25, 27, but we needn't get into them.

The second goal that we have mentioned, that of encouraging investment by the licensee, is the relevant one in this case. Angiotech does not manufacture or coat stents itself. It depends on its licensees to develop the product (that is, the coated stents) and obtain the Food and Drug Administration's approval so that the product can be marketed. If, therefore, Cook or BSC were essentially interchangeable, or one were clearly a superior developer to the other, we would expect Angiotech to grant an exclusive license to one of them. Probably the reason it did not is that the two firms use different coating methods, requiring each to obtain separate approval from the FDA before being permitted to market a drug-coated stent in the United States. When the contract was made, and indeed to this day, neither Cook nor BSC had yet obtained FDA approval. Their products are still being tested for safety and efficacy. Angiotech would not have wanted to risk betting on the wrong horse — granting BSC an exclusive license, for example, when Cook's stent might turn out to be the only drug-coated stent that the FDA would approve, or might be approved earlier than BSC's product, or might prove to be the superior product.

At the same time — and here we approach the crux of the appeal — Cook and BSC might be reluctant to accept nonexclusive licenses. To obtain FDA approval for a new drug or medical device (a paclitaxel-coated stent is what the FDA calls a "combination product" and is assigned to the division of FDA that handles devices) requires a substantial investment, which the manufacturer of the device might have difficulty recouping if it faced competition from another licensee. With Angiotech reluctant to grant an exclusive license and Cook and BSC reluctant to settle for nonexclusive licenses, coexclusive licenses were a natural compromise.

The compromise might be undone by assignment or sublicensing. Suppose Cook fell behind BSC in the race to develop an approved and marketable paclitaxel-coated stent and tried to recover the lead by assigning its license to a firm with greater resources or other advantages that would enable it to overtake BSC. The effect would be to confront the latter with more competition than it had reckoned on when it took out its license from Angiotech. Hence the bar on assignment without the permission of the other licensee (plus Angiotech, though that is not a factor here). Although competition is generally a good thing, there is no argument that Angiotech would have violated antitrust law or been guilty of patent misuse had it granted an exclusive, nonassignable license. And so we cannot see how Angiotech's action in granting two licenses and forbidding the licensees to increase the number of competitors by means of assignment or sublicensing could raise an antitrust or patent-misuse issue. Notice that the licenses do not forbid the acquisition of the licensee by another firm — to which as an affiliate the licensee could grant a sublicense without obtaining the permission of the other parties to the contract with Angiotech — that might be a more formidable competitor of the other licensee.

Angiotech granted the coexclusive licenses in 1997. Four years later Cook made a contract (actually five simultaneous contracts, but that is a detail of no legal significance) with Advanced Cardiovascular Systems, Inc. (ACS), a manufacturer of medical devices and a subsidiary of Guidant Corporation. Under the contract, Cook is to purchase stents from ACS, coat them with paclitaxel, and sell the coated stents back to ACS for resale to hospitals and other purchasers of medical devices. The price received by Cook for the stents that it sells to ACS is to be one-third of the resale price charged by ACS. Before reselling the stents to hospitals or other users of stents, ACS is to mount them on a catheter (which it manufactures) for inserting the balloon and stent into the artery. So what it is selling is really a stent system rather than merely a stent. The stent system is called "ACHIEVE," an ACS brand, though Cook's name will also appear on the package. The contract requires ACS to obtain the regulatory rulings necessary to enable ACHIEVE stent systems actually to be sold, such as approval by the FDA.

BSC argues and the district judge found that the transaction between Cook and ACS is a de facto assignment and so violates the contract because BSC did not consent. (ACS like Angiotech is not a party to the suit — which is a surprise, as one might have expected BSC to join ACS as a counterdefendant on the theory that ACS committed the tort of interference with BSC's contract with Angiotech and Cook.) Cook argues that since the finding was made in a summary judgment proceeding rather than after a trial, we should review it de novo, that is, without according any deference to the district judge's ruling. This is a strange argument. A judge makes findings in a trial or other evidentiary hearing; a grant of summary judgment is a determination that there are no triable issues. So one would expect Cook to be asking not for de novo review of the district judge's findings, but for a trial. One possibility is that Cook thinks (incorrectly, as we'll see) that the question whether the de facto assignment violated the contract is a pure question of law. Another is that Cook consented to have the district judge decide the question of contractual violation on the basis of the pleadings, the briefs, and the limited discovery that had been conducted before the motion for summary judgment was made. Both inferences are supported by Cook's opposition to BSC's completing discovery.

When litigants waive trial, and ask the judge to decide the case as if the record compiled in the pretrial proceedings were a trial record, appellate review is as of findings made after a trial, not as of a grant of summary judgment. "[S]ometimes both parties move for summary judgment because they do not want to bear the expense of trial but instead want the trial judge to treat the record of the summary judgment proceeding as if it were the trial record. In effect the judge is asked to decide the case as if there had been a bench trial in which the evidence was the depositions and other materials gathered in pretrial discovery." May v. Evansville-Vanderburgh School Corp., 787 F.2d 1105, 1115 (7th Cir.1986); see also Hess v. Hartford Life & Accident Ins. Co., 274 F.3d 456, 461 (7th Cir.2001). It is true that the mere fact that cross-motions for summary judgment are filed, as was done in this case, does not operate as a waiver of the right to a trial. Miller v. LeSea Broadcasting, Inc., 87 F.3d 224, 230 (7th Cir.1996). But though there was no explicit waiver, it is reasonably clear that Cook didn't want a trial (more precisely, it either wanted a trial limited to the summary judgment record or thought the issue of contractual violation could be resolved without a factual record) and so waived its right to a (fuller) trial. In its opening brief on appeal, Cook belatedly requested a trial if summary judgment in its favor was rejected. BSC replied that the demand comes too late — and Cook apparently agrees, for its reply brief is silent on the point: a second waiver.

The interpretation of a written contract, when no extrinsic evidence (evidence other than the contract itself) is presented, is treated as an issue of law, and thus is decided by the appellate court de novo, that is, without giving the trial judge's interpretation any special weight. But the rule is otherwise when as in this case the resolution of the interpretive issue requires a comparison of written documents and thus an inference from multiple pieces of evidence, a traditional task of a finder of fact, rather than requiring merely "gazing" at a single document. Coplay Cement Co. v. Willis & Paul Group, 983 F.2d 1435, 1438 (7th Cir.1993). (The comparison required in this case is between the three-cornered Angiotech licensing contract with Cook and BSC and Cook's contract with ACS.) In such a case the ruling of the trier of fact is reversible only if clearly erroneous. E.g., In re Modern Dairy of Champaign, Inc., 171 F.3d 1106, 1109 (7th Cir.1999); Glass v. Kemper Corp., 133 F.3d 999, 1001 (7th Cir.1998). No matter. The district judge's ruling was not erroneous at all, and so the precise standard of review is unimportant.

Had the contract between Cook and ACS provided that Cook would license to ACS the use of paclitaxel in the ACHIEVE stent system, this would have been an assignment or sublicense in violation of the anti-assignment clause; and as that part of the contract was intended, in part anyway, to protect BSC, BSC has a right to enforce it. (It would have a right to enforce it, as a third-party beneficiary, even if it were not a party to the contract. See A.E.I. Music Network, Inc. v. Business Computers, Inc., 290 F.3d 952, 955-56 (7th Cir.2002); Vidimos, Inc. v. Laser Lab Ltd., 99 F.3d 217, 219-20 (7th Cir.1996); Postlewait Construction, Inc. v. Great American Ins. Cos., 106 Wash.2d 96, 720 P.2d 805, 806-07 (1986).) This conclusion would be unaffected if besides licensing the use of paclitaxel to coat ACS's ACHIEVE stents, Cook had agreed to do the coating itself (as it did agree in its "sale" contracts with ACS). Cook would still be licensing the distribution and sale of paclitaxel in stents to ACS, a firm with which Cook is not affiliated.

This means that Cook's defense to BSC's charge of breach of contract hinges on the fact that Cook is to sell the coated stents to ACS rather than assigning its patent license to ACS. Cook describes the "sale" as the exercise of the right granted to it by Angiotech to distribute and sell Angiotech's patented product. But the sale of ACS's stents back to ACS has no commercial purpose or substance; it is merely a device for defeating the anti-assignment clause. (By the way, no supposed public policy against anti-assignment clauses is invoked by Cook. Compare Bank of America, N.A. v. Moglia, 330 F.3d 942, 947-48 (7th Cir.2003).) Suppose a Mr. Guidant asks a Mr. Cook to paint his house. And Cook says, fine, but let's do it this way: you sell me your house, I'll paint it (supplying both the paint and the labor to apply the paint to the house), and then I'll sell the house, painted, back to you. That would make no commercial sense, and so one would delve for an improper motive. Confronted at argument by the housepainting hypothetical, Cook's lawyer was unable to distinguish it from this case.

It is obvious what is going on. Cook wanted to improve its competitive posture vis-à-vis BSC by obtaining the resources of a firm that had a better stent than Cook itself. It could have done so without breaking its contract, by affiliating with ACS; affiliation, as by the sale of Cook to Guidant (a transaction actually contemplated at one point, as we'll see), would have allowed Cook to grant a sublicense to ACS. This would not have been an evasion of the three-cornered contract, because the sublicensing of an affiliate is authorized by the licenses. It might appear to be a case of taking advantage of a loophole, but maybe not; it is more costly to merge with another firm than to execute an assignment or the series of five contracts that in substance were an assignment. We know it's more costly because after the district court enjoined the assignment, Cook entered into merger discussions with Guidant, which failed; had they succeeded, the cost to Guidant would have been in the neighborhood of $3 billion.

But while a merger plus an assignment or a sublicense differs in substantial and not merely formal respects from an assignment, the five contracts between Cook and ACS, realistically treated as one, differ from an assignment only in formal, in the sense of economically empty, respects. See In re Shulman Transport Enterprises, Inc., 744 F.2d 293, 295 (2d Cir.1984); cf. Grojean v. Commissioner, 248 F.3d 572, 574 (7th Cir.2001); SEC v. SG Ltd., 265 F.3d 42, 46-47 (1st Cir.2001). Not that sale-and-leaseback arrangements, which the Cook-ACS transaction superficially resembles, are characteristically devoid of economic substance, though they can be, as in Coleman v. Commissioner, 16 F.3d 821, 826-27 (7th Cir.1994), since a common purpose is to give the buyer-lessor tax benefits. See, e.g., Solargistic Corp. v. United States, 921 F.2d 729, 730 (7th Cir. 1991). But emphasis falls on "superficially." Cook didn't buy stents from ACS and then lease them back; it bought them from ACS and then resold them to ACS, the only purpose of the transaction being to transfer Cook's patent rights to ACS in circumvention of the anti-assignment clause.

A further issue discussed in the briefs is whether Cook also violated the clause in the contract that makes it responsible for obtaining the necessary regulatory approvals for the sale of a paclitaxel-coated stent. BSC argues that Cook shifted that responsibility to ACS, and Cook replies that it retained ultimate responsibility and anyway that Angiotech hardly cares whether Cook or ACS takes the laboring oar in obtaining the necessary approvals. Cook is probably right; but the debate is beside the point. The significance of the provision in the contract between Cook and ACS that assigns to ACS the task of obtaining regulatory approvals is merely as further evidence that the contract actually assigned the sale of paclitaxel to ACS, because ACS did everything except the coating to bring the paclitaxel to market.

So Cook broke its contract with BSC and the next question is the propriety of the relief granted by the district court. As BSC points out, it would be very difficult for a court to estimate the damages that it has incurred as a result of Cook's breach. The reason is that neither party has yet obtained the FDA's approval to sell its product. The de facto assignment to ACS undoubtedly gave Cook a leg up, but to translate this insight into a dollar amount of lost expected profits to BSC is impossible. No one knows whether or when, as a consequence of the assignment, Cook will obtain FDA approval or how well its product will succeed in the market. No one knows whether or when BSC will obtain approval of its paclitaxel-coated stent and how successful that product will be in the market. There is injury to BSC in a probabilistic sense — enough injury to establish standing and entitle it to relief of some sort — but whether the injury will prove to be $1 or $100 million is unknown and unknowable.

When a breach of contract is proved but damages cannot be estimated with reasonable certainty, the plaintiff is entitled to an injunction. But the injunction the district court issued goes too far. It not only forbids Cook to perform its contract with ACS or otherwise to violate its contract with Angiotech and BSC; it also provides that "no information, data or technology generated or gathered in connection with the ACS deal shall be used for any commercial purpose, including the purpose of obtaining regulatory approval to sell paclitaxel-coated stents in the United States or elsewhere." The passage that we have italicized violates the principle that in determining the appropriate scope of an injunction the judge must give due weight to the injunction's possible effect on innocent third parties. Association of Community Organizations for Reform Now v. Edgar, 56 F.3d 791, 797 (7th Cir. 1995); Kasper v. Board of Election Commissioners, 814 F.2d 332, 340 (7th Cir. 1987). In this case they are the sufferers from atherosclerosis who might benefit from a device that prevents restenosis. Those effects must be balanced against the harm to BSC from narrowing the injunction by lancing the italicized phrase — but as it happens that harm is zero in a legal sense of "harm," which differs from harm in the lay sense. The difference is brought out in the legal slogan damnum absque injuria, i.e., harm without a legally cognizable injury.

What would harm BSC in a legal sense would be the competitive impact on it of having to compete with the Cook-assisted ACHIEVE stent system; it would not be the regulatory approval of that sale if the injunction against sale remained in place so that Cook could not use the approval to enable ACS to sell ACHIEVE in competition with BSC. When pressed at argument BSC's lawyer named only one harm to BSC from the grant of such approval itself — and that a shocker: that the FDA might discover in the course of considering Cook's application (prosecuted on its behalf by ACS) for approval that paclitaxel-coated stents are harmful or ineffective, and that the discovery would hurt BSC, whose own product is also paclitaxel-coated. Indeed it would — and should. We shall therefore modify the district court's injunction by striking the italicized phrase.

Should Cook obtain the FDA's approval before BSC does, it will, as we have been at pains to emphasize, still be enjoined from selling its product though authorized by the FDA to do so. But at that point we imagine that the parties will be able to negotiate the dissolution of the injunction on terms that compensate BSC for having been beaten to the punch. Of course, this suggests that Cook will derive a commercial advantage from being able to continue seeking the FDA's approval to sell a product that Cook is enjoined from selling. But that advantage, which will in any event be shared with BSC in the terms of settlement, seems slight in relation to the social costs of delaying the process of FDA approval. Indeed, should the negotiations we envisage (on the assumption that Cook obtains the FDA's approval before BSC does) fail, the district court might well decide to modify the injunction so that people suffering from atherosclerosis can obtain the benefit of Angiotech's technology at the earliest possible opportunity. See Fed. R.Civ.P. 60(b)(5); Protectoseal Co. v. Barancik, 23 F.3d 1184, 1185-87 (7th Cir. 1994); In re Hendrix, 986 F.2d 195, 198 (7th Cir.1993); Bellevue Manor Associates v. United States, 165 F.3d 1249, 1255-56 (9th Cir.1999); Alexis Lichine & Cie v. Sacha A. Lichine Estate Selections, Ltd., 45 F.3d 582, 585-86 and n. 2 (1st Cir.1995). While Cook errs in suggesting that 35 U.S.C. § 271(e)(1), which broadens the experimental-use defense to patent infringement as explained in SmithKline Beecham Corp. v. Apotex Corp., 247 F. Supp. 2d 1011, 1018 (N.D.Ill.2003), is a defense to a breach of contract suit (the provision is expressly limited to patent infringement suits), the section does reflect a policy of allowing the use of patented technology to obtain regulatory approval of noninfringing technologies. There is no suggestion that in using paclitaxel to obtain FDA approval Cook would be infringing Angiotech's patent. And so the policy that we have just mentioned is something the district court could and should consider if it is ever asked to modify the injunction to enable paclitaxel to be made available to the public.

One loose end remains to be tied up. In the course of the limited discovery that took place before the district court granted summary judgment in favor of BSC, Cook inadvertently turned over to BSC documents that were privileged. The judge held that the disclosure waived privilege, but he did not consider the documents in making his decision, and BSC has since returned them to Cook. Cook argues that the judge erred, but acknowledges that the argument is moot unless we reverse the finding of liability and remand for proceedings in which the documents might unless privileged be used by BSC as evidence against Cook. The documents relate to liability rather than to relief and so their admissibility is indeed a moot point.

The judgment, as modified to narrow the injunction, is affirmed.