Patent exhaustion and open source

Please consider subscribing to LWN

Subscriptions are the lifeblood of If you appreciate this content and would like to see more of it, your subscription will help to ensure that LWN continues to thrive. Please visit this page to join up and keep LWN on the net.

When patents and free software crop up together, the usual question is about patent licensing. Patent exhaustion — the principle that patent rights don't reach past the first sale of a product — is much less frequently discussed. At FOSDEM 2019, US lawyer Van Lindberg argued that several US court decisions related to exhaustion, most of them recent but some less so, could come together to have surprising beneficial effects for free software. He was clear that the argument applied only in the US but, since court systems tend to look to each other for consistency's sake, and because Lindberg is an engaging speaker, the talk was of great interest even in Brussels.

[Van Lindberg]

A patent is a limited legal monopoly granted to protect an invention, giving the holder the right to exclude others from using, making, selling, and importing the invention (including things that embody the invention) for a fixed period of time. Much has been said and written over the years about the extension of patents to cover ideas that are expressed in software, but software patents are definitely with us at the moment.

There are, however, a number of limitations on the rights that a patent grants. One of these is patent exhaustion, which protects the ability of those lawfully in possession of goods embodying patents to use, sell, or import those goods without interference from the patent holder. Exhaustion prevents the patent holder from profiting more than once from the sale of any particular item; in Lindberg's words, as soon as the patent holder puts something "into the stream of commerce", the patent rights are exhausted. If Alice holds a patent for an invention embodied in a widget, and she sells a widget to Bob, then Bob is protected against accusations of patent infringement because he acquired the widget from the patent holder. If Bob sells his widget to Carol, she is similarly protected; not because she has licensed the patent from Alice, but because Alice's patent interest in that widget was exhausted by that first sale to Bob.

Naturally, over the years enthusiastic patent holders have tried a number of tricks to do end-runs around patent exhaustion, but often the courts have knocked them back. Lindberg outlined the limiting principles provided by five such cases that clarify the extent of patent exhaustion and underlie his surprising argument:

  • Exhaustion results from any authorized transfer: Lexmark sells printers, said Lindberg, and the company really wants you to buy its expensive ink cartridges. So Lexmark sold ink cartridges embodying patented ideas on the condition that empty cartridges must be returned to them. Someone else acquired empty cartridges, refilled and sold them, and got sued for patent infringement. In Impression v. Lexmark, the US Supreme Court said that wasn't going to fly. There might be a contractual issue between Lexmark and the original purchasers of the cartridges, but the patents were exhausted by the sale of the cartridges, and patent law could not be used to pursue the reseller.
  • Exhaustion applies to both system and method patents: There is more than one kind of patent. Among those types are systems patents, each of which covers a novel "product, device, or apparatus", which is to say, a tangible item. There are also method patents, each of which covers a novel "series of acts or steps", which is to say, a way of doing things. In Quanta v. LG it was argued that a method patent couldn't be sold embodied in an item in the same way that a systems patent could, and so the patent could not be exhausted by the sale of an item. The US Supreme Court said that "a patented method may not be sold in the same way as an article or device, but methods nonetheless may be 'embodied' in a product, the sale of which exhausts patent rights".
  • Exhaustion applies even when there is an express reservation of patent rights: Some people, said Lindberg, have tried to exclude patent rights from the sale. There was a Xerox PARC license that was essentially the BSD license with added terms that limited the license only to copyright, excluding any license of patent rights unless the rights-holder explicitly added their name. But in Impression v. Lexmark, the US Supreme Court said that "this court accordingly has long held that, even when a patentee sells an item under an express restriction, the patentee does not retain patent rights in that product".
  • Exhaustion applies to any authorized transfer, not just sales: Others have tried to argue that if they give something away, instead of selling it, then patent rights are not exhausted. In LifeScan Scotland v. Shasta Techs, though, the US Court of Appeals for the Federal Circuit was particularly unimpressed by this argument, saying that "a patentee cannot evade patent exhaustion principles by choosing to give the article away rather than charging a particular price for it". Lindberg noted that this is directly applicable to free and open-source software (FOSS). As he amusingly summarized it, if you choose to transfer your widget for zero dollars, you can't come back later and complain you didn't charge enough and should therefore still have patent rights to enforce. So even if we didn't have Jacobsen v. Katzer telling us that there is an economic benefit associated with having people use your free software, the zero-cost nature of free software wouldn't prevent the doctrine of patent exhaustion from applying.
  • Exhaustion applies to foreign sales: Also from Impression v. Lexmark, an authorized sale outside the US, just as one within the US, exhausts all rights under the Patent Act.
  • Exhaustion prevents patent assertions against authorized recipients of FOSS: In Cascades Computer Innovation v. Samsung Electronics, the dispute arose because Cascades had licensed some patents to Google for use in certain ways inside the Dalvik virtual machine with a condition that Google could only use the license for Google products. One such product was the Android Open Source Project (AOSP). Samsung took the AOSP code, compiled it, and distributed it on its phones. Cascade sued, but the Northern District Court for Illinois, according to Lindberg, said "no", because once you've put something into the stream of commerce, your patent is exhausted. The AOSP came from a producer with a patent license; the fact that Google chose not to charge money for its product doesn't preclude exhaustion.
  • Code written by non-patent-licensees can still exhaust the patent if it is distributed by a licensee: Lindberg was quite coy about the fifth case, Intel v. ULSI, when he introduced it earlier in the talk. Coming back to it, he described it as the kicker, the one with the twist, even though coming from 1993 it's by far the oldest of the judgments. In this case, HP was given a license by Intel to be a foundry for certain computer chips: to manufacture and sell them to third parties. Another company, ULSI, designed its own, similar chip, and asked HP to manufacture it. HP did so, at which point Intel sued ULSI for infringing Intel's patents, as ULSI had obtained no license from Intel. The US Court of Appeals for the Federal Circuit held that because HP had manufactured the chips, and because at the time it did so it held a license to the patents, no infringement had occurred; any sales of ULSI chips were lawful and thus exhausted those patents. Code is a good, just like any other product, as Lindberg confirmed in response to a later question, so the passage of code through the hands of a patent licensee effectively "sanitizes" the code with respect to those patents, exhausting them in the process.

Having established his list of principles for patent exhaustion, Lindberg described a hypothetical scenario. Suppose that Alice made a chip that embodied a patent owned by Bob. Alice had no license to do it, she just created this chip, ran off millions of copies, and they got embedded in phones; that sort of thing, said Lindberg, happens all the time. Carol realizes this chip is embedded in phones which can be bought from Bob's Phone Shop. She buys a lot of these phones, cracks them open, extracts Alice's chips, and makes a nice business out of reselling them. Bob sues Carol for violating his patents, saying that he didn't give Alice a license for the chips that she made and Carol bought. Carol, however, argues that she got the phones from Bob, and using the principles above, wins. It turns out, said Lindberg, that the analysis is no different if it's a piece of software than if it's a chip in a phone.

We in the free software world have repositories, distributions, and mirrors; copies of source code are hosted by companies willy-nilly. Suppose that some company had mirrored a copy of a Linux distribution, with its thousands of constituent programs, each of which might embody one or more patents. Then that same company, because it is an authorized licensee for such of those patents as the company itself either held or had a right to use (by virtue of being in one or more patent pools or cross-licensing arrangements), would have exhausted those patent rights with respect to that software. Lindberg did add a caveat, however: courts frequently try to avoid surprising outcomes, therefore a court might follow the argument but decide not to allow it anyway.

At this point, Lindberg reminded attendees that Microsoft bought GitHub. After a short pause, the entire room, with a large proportion of lawyers in the audience, giggled, a sound that can only be described as chilling, then applauded. He then went further and proposed an N-way merge across copies of code bases sanitized by different distributors with respect to their different patent portfolios, to create code bases that are exhausted with respect to all patents that all those various distributors are authorized to use.

This was a fairly difficult talk to follow, and it's not an argument I've heard before. But the audience reception was fairly friendly; there were a couple of detailed legal questions about the implications of other judgments, but Lindberg didn't seem to feel they were fatal. I hope to hear this argument a lot more in the future because, if it works, it bodes well indeed for controlling software patents.

The original talk can be seen and heard here.

[We would like to thank LWN's travel sponsor, the Linux Foundation, for travel assistance to Brussels for FOSDEM.]
(Log in to post comments)