Google v. Oracle argument preview

Chicago-Kent College of Law 2020 Supreme Court IP Review: Google v. Oracle

Copyrightability issue: Pam Samuelson: Supreme Court has 2x decided that functionality defeats ©ability for certain words/symbols when they constitute a system/method for accomplishing a useful result. Paris v. Heximer, a claim of infringement over D’s copying of a symbol system for ID’ing different types of buildings for fire insurance maps. Baker v. Selden too. Didn’t matter how original symbols/words were or what other choices were possible at the time of creation. Promotes progress of knowledge by allowing authors to build on existing knowledge and promotes useful arts by preventing © from being misused to give protection to patent-like inventions that didn’t satisfy patent standards. Dozens of cases on how shorthand systems, tax record systems, games are unprotectable even when embodied in a © work. Codified in §102(b). Codification came in order to deal w/fears about protecting computer programs as literary works, per legislative history; Fed. Cir. instead ignored every word in §102(b) except “idea” and ignored legis. history of concern for overprotection.

Fed Cir never identified the relevant work of authorship or asked what the scope of protection available to the work in light of Baker & its codification in §102(b). The relevant work is Java SE, which has the API and numerous other elements of the Java technologies created by Sun & its engineers. The Q is whether the specific declarations Google interspersed in 15 million lines of code are w/in the scope of protection for Java SE. Judge Alsup made findings of fact that these were methods/systems, consistent w/9th Cir. law which has twice held that program interfaces that enable interoperability are unprotectable procedures under 102(b). Sun—developer of Java—participated in an organizational brief in support of Borland in Lotus v. Borland, committing to proposition that interfaces that enable interoperability should be treated as a system.

Interfaces are fundamentally distinct from implementations: everyone has to write their own reimplementations, but reimplementations themselves are legitimate, which promotes interoperability and competition.

Merger doctrine leads to the same result. Once you decide to adopt a certain system, that constrains potential implementations. Many cases involving 102(b) thus use merger as an alternative argument.

Anne Capella (Weil): Programmer; represented computer science professionals to talk about the underlying technical aspects. APIs are tools for app developers to write applications. They are the audience for the APIs. Try to make them easy to use and intuitive. Also explained: what is declaring code? May look like a collection of random words, but there are a lot of subjective decisions that have to go into designing those and that are not completely dictated by function. There is original expression. Like a haiku: there are constraints on the format, but also subjective decisionmaking and creativity. API designer that would design declaring and implementation code would be balancing functionality, flexibility, simplicity and trying to make those attractive to app developers, requiring selection of which function to present, how to organize. Not just a filing cabinet but hierarchies that are important as to their locations. There are good APIs and APIs that are more attractive than others; Java was known for being well designed and intuitive which made it attractive to app developers. Better APIs have better expressions for the same function, which are not dictated by the functions themselves.

You can also write your own programs to provide the same functionality that wouldn’t have to use the same name, organization, design choices for what would be in the declaring code. [Which seems a lot like games and shorthand systems.]

Interoperability: there was no technical need to use the declaring code. Programs written for Java SE can’t be used on Android w/o modification and vice versa. Not all of the APIs are available in both. There wasn’t a need to use them to create a new Java platform. Google could have written its own declaring code but wanted the intuitive code and wanted to attract programmers to their own platform. Programmers want to protect expression in declaring code. If it’s protected, people might try to hold interfaces proprietary, but Java had a license that Google chose not to use.]

RT: (based on this amicus)

At its core, the fair use argument for Google is about why fair use is a multifactor test and why there are different kinds of fair uses. I will touch on deference to the jury though I have plenty of other thoughts on that for the question period.

Fair use has four nonexclusive statutory factors: the purpose of the use, the nature of the accusing work, the amount of the original work taken, and the effect on the market for the original work.

Beginning as usual with factor one: The extent to which a new work has a new meaning, message, or purpose—transformativeness—is often and rightly prioritized in the fair use analysis. But what constitutes transformativeness is often contentious. Here, the new purpose of Google’s new code implementing the declarations was the creation of a new computing environment in which Java programmers could readily create programs on multiple platforms, which required the use of limited portions of highly functional declarations. This “technical interchange,” is a specific kind of creativity-promoting transformativeness.  It’s about the programmers, not whether the program is “run once run anywhere,” which is a red herring. This type of purpose has been recognized as transformative because of its role in furthering competition and innovation. A computer interface supports the creation of other creative works, and in such situations, it is important to avoid locking in third parties to specific platforms.

By contrast, the Federal Circuit reasoned that factor one weighed against Google because the intrinsic purpose of the parties’ declarations was the same. But that reasoning creates a per se rule against copying any interface elements or code in a computer program, and fails to appreciate the interaction of factors two and three with transformativeness. The Federal Circuit’s simplistic rationale suggests that a law review article that quoted another law review article for support would infringe the earlier article, because they both had the same “intrinsic purpose.” Without assessing the role of the copied material in the parties’ works, this is a mistake.

Factors two and three of the fair use test help define the boundaries of technical interchange fair use. The highly functional nature of the copied declarations and the limited amount of the overall Java SE work used, consistent with industry practices, support the conclusion that Google’s use was a transformative use that served copyright’s basic goal of encouraging creation of new works.

As to the nature of the original work: The jury heard evidence that the declarations and classes of the Java SE API were functional, not merely in the way that all computer code performs a function, but specifically in that these particular declarations perform their mini-duties in noncreative ways. The Federal Circuit acknowledged the thinness of the copyright, but it failed to put this factor in proper context and thus to give it actual weight, instead reasoning that the second factor did not matter to fair use determinations generally, a proposition for which it cited only cases involving highly creative works like songs. But there are four factors because various fair uses can be fair for different reasons, and the factors may thus differ in relevance across types of cases.

Here, the reasons that factor two favored Google are highly relevant to the overall fair use analysis. No matter how much work and how many choices went into producing Java SE, the highly utilitarian nature of declarations means that copyright grants them thin scope at best. This thinner scope of protection naturally leads to a broader scope for fair use, especially in conjunction with factor three.

On factor three, amount, the question is not whether, in judicial hindsight, the defendant took more than absolutely necessary, but whether the amount taken was “reasonable in relation to the purpose of the copying.” As the jury heard and evidently credited, Google took an amount from Java SE considered in the industry to be reasonable in light of its purpose of developing new, compatible works.

One aspect of the Federal Circuit’s error was its emphasis on the 37 packages that Google copied in “entirety.” In a fair use analysis, it is vitally important to identify the allegedly infringed work or works. As the statute commands, the proper inquiry considers the amount taken “in relation to the copyrighted work as a whole.” Without careful attention to the boundaries of works, especially unfamiliar works such as software, plaintiffs can manipulate their claims to artificially increase the relative size of what was taken. As Justin Hughes has written, “If our goal is to create special incentives for the building of houses, we do not necessarily need special incentives for the making of bricks or the mixing of mortar. . . ..”

So what was the work at issue? Both the copyright registrations and industry practice were clear: the work is Java SE, which had about 5 million lines of code; Android was three times bigger. Google’s implementation expressing the declarations at issue totals about 11,500 lines. As a quantitative matter, what was copied was a rounding error. Yet the Federal Circuit characterized Google as “taking a copyrighted work verbatim,” as if the fractional part were the whole.

As the district court detailed, the jury could have found that Google used highly limited portions of Java SE’s APIs as part of an innovative mobile device platform. The jury heard testimony that portions of Java SE that Google reimplemented preserved consistency of use within the larger Java developer community. What’s more, a large number of books have been published that set forth the Java SE API, in whole or in part, including the declarations and class structures. Numerous witnesses, including Sun’s then-CEO who was there when it developed Java, testified that Google behaved according to industry practice: write your own implementing code but APIs are for everyone to use.

Where others were freely copying large parts or the entirety of the API, the jury could have found that Google’s far more limited copying strongly favored Google. In this way, the thinness of the copyright—factor two—interacts with factor three: even if APIs as a whole cross the line into copyrightability, Google should be able to use declarations (and the organization they necessarily reflected) that were reasonably necessary to pursue its legitimate goal of enabling the creation of an environment accessible to Java programmers.

Which leads to factor four: By downplaying the relevance of the nature of the work and the amount taken, the Federal Circuit fell into the well-known trap of circularity: reasoning that, because Oracle could have charged a license fee for this type of use if fair use were unavailable, Oracle therefore suffered cognizable market harm.

Because such claims can be made for any fair use, which by definition is not paid for, this reasoning cannot distinguish fair and unfair uses. In some situations involving traditional creative works with “thick” copyrights, experience and normative commitments to free speech break the circle: the markets for criticism, educational uses, and parody are not legitimate even if a specific copyright owner evinces a willingness to license particular instances.

When it comes to software, the benefits of interoperability and allowing third parties to avoid lock-in to particular platforms similarly explain why licensing interface reimplementation in newly written code is likewise not a legitimate copyright market.  Here, factors two and three can help identify when crediting claimed market harm would be inconsistent with copyright’s overall balance between past and future creators.

In reversing the jury’s verdict, the Federal Circuit wrongly relied on the circular claim that Google could have taken a license for the precise use it made and that this possibility established market harm. But the evidence did not establish harm to the market for “the work,” Java SE, or even the existence of a market for a limited number of highly functional declarations.

A thin copyright for software, including Java SE, provides software copyright owners with meaningful protection against copying of significant amounts of expression, but meaningful protection does not require the expansive rights that the Federal Circuit granted.

Elizabeth Brannen, Stris & Maher: Nothing is free. [Other than this webinar?]

What Google copied was expression that was popular with Java developers and that cost lots of resources to create. Those packages covered 6000 separate methods that Google didn’t have to write itself. Those packages are in executable form on countless Android devices.

Interfaces v implementation is a false dichotomy. We need to look at what was copied and was it expression. The haiku is a good analogy. Best interfaces are concise and intuitive, which are hallmarks of good writing in any context. Google copied b/c 6 million programmers knew and liked the Java interfaces to call the prewritten methods. It copied what it did because of the expressive value of that code. Most computer code is zeros and ones. Google copied what’s meaningful to developers, the expressive part. It didn’t copy for technical reasons. Wrote its own versions for many Android declarations. Copied not b/c of technical impossibility but b/c didn’t want to require Java developers to learn new calls, so it helped itself to the ones they already knew and liked. Google made a lucrative business decision. To call the © thin is to say functional works should receive second-class protection. [We have thin and thick © for a reason—of course it’s about scope. It’s not second-class to rece] Google didn’t transform. It’s not fair to replace and supersede the original.

Transformative use changes something about the message/content—parody, commentary. Google killed Java SE for mobile. There was unrebutted evidence that they designed Android to be incompatible w/Java. No court has found fair use where there was an incompatible software product. [Connectix found fair use where there wasn’t full compatibility, as here.] Google took a shortcut and made money. Fair use would undermine the incentive to create new works. People only take licenses because they have to. Expressive aspects of code need © to incentivize investment and making code available under unrestrictive licensing terms.

Ned Snow, South Carolina Law: The only Q the Supreme Court will care about: In the late 1700s, did the issue of fair use determine whether a defendant had infringed the legal right of copyright? The test isn’t which court was the first to decide, equity or law. Not whether fair use more frequently arose in a court of low or equity; that depends on what remedy is sought (injunctive relief or damages). Also not whether a ct of equity would have granted a D’s request for a court of law. Not whether a court of equity states that judges should detrmine the issue. He argues that two cases determine the answer. [This is no way to run a railroad, which is not Prof. Snow’s fault at all, but his explication trying to extract fair use from a pre-modern system convinces me of the disutility of the enterprise.]

Sayre v. Moore, Lord Mansfield, fourteen lines of legal analysis. Mansfield recognizes value in correcting errors in faulty sea charts. Charges the jury to find whether D corrected errors or servilely copying: value and purpose of D’s use. Points to allegation that alterations are “very material” which points to transformation. Also encourages jury to consider whether D’s work enables or worsens navigation, which points to nature of work.  If you think so, he tells the jury, you find for the D; if you find mere servile imitation, you should find for P. Shows that jury determined the infringement of the legal right. Neither Mansfield nor jury was concerned with independent creation, or fact/expression dichotomy. Lockean labor theory/no reason to think that factual works had less protection than expressive works. Even if we inferred that, the inference would support using fair use which considers fact/expression in factor two.

Next question: what is the review standard for mixed questions of law and fact. Descriptive argument: what’s going on? Take a raft of case-specific historical facts, balances them, and produces a result. That’s pretty factual sounding, per J. Kagan in US Bank v. Village at Lakeridge.  When an issue falls b/t pristine legal standard and simple historical fact, standard of reivew reflects which judicial actor is best positioned; juries have a variety of life experience, suggesting deference to them, though judges can issue general principles.

William Jay, Goodwin Procter: Bottom line: accepting Google’s argument in its strongest form, fair use is a four factor test and not only is each factor grounded in the facts, but the balancing is itself a factual question. That’s the key assertion here. If the Ct agreed, it would circumscribe JMOL review, but would also make summary judgment all but impossible even on an undisputed factual record. Skeptical the Court will go there in its strongest form. There was a jury trial! So this is a case about the appellate standard of review. Oracle does say there was no right to a jury trial anyway, so the 7th Amendment is irrelevant, but at the end of the day the Fed Cir was reviewing a JMOL verdict after a jury trial. Everyone pretty much agrees that the standard of review is in a sense de novo, but de novo review of what? Oracle’s position is grounded in Harper & Row, which was a bench trial; J. O’Connor said that the appellate court didn’t need to remand but could conclude no fair use as a matter of law.

How should that work here? At a minimum, jury’s job is to find historical facts: how much of the work was taken? Those historical facts need to be taken in the light most favorable to the prevailing party. But many Qs are Qs of degree, along with the final balancing of factors. Is that final balancing factual? Harper & Row didn’t remand for a new finding, but instead said that it was important to have historical facts determined and then the reviewing court could make its own determination. The remainder of the analysis is the province of the courts in a non-bench trial case.

If Google is right, there’s no work for judges to do after jury verdict, b/c it’s always possible that some kind of weighing favors the verdict. But the Court likely won’t reach that; Oracle agreed to a jury trial and there’s no need to decide how much of a right to a jury trial there is to decide the standard for JNOV.

Tomas Gomez-Arostegui, Lewis & Clark Law: 7th Amendment test is historical, turning on English authorities circa 1791. But Ct often not clear about what precisely it seeks from the record. Analogues work but how analogous they need to be is unclear. Feltner for example didn’t deal w/statutory damages. We’re looking for “established practice” but neither of those words are defined—Feltner cited two cases in which juries awarded nominal damages.

© litigation before 1800 was mostly in equity; ordinary damages were difficult to prove and even penalties were hard to obtain; there was no provision for discovery in law courts and P & D couldn’t even testify. Equity was easier! Chancellor could determine law and facts. Equity wasn’t obligated to send © suits to law. If Chancery did push a case to law, it was typically for a sticky legal interpretation Q. If they did go to law, there were lots of issues that were sent to the jury.

What about fair abridgement? 18th-c analogue to fair use? Oracle stresses it’s an equitable doctrine, w/no right to jury, and also argues that it’s not the same as fair use. Fair abridgement raised in equity from 1680s through 18th c, but no case at law before 1800 of which he was aware adjudicated fair abridgement. That doesn’t make it a purely equitable doctrine, esp. since it came from a statute that would apply equally at law and equity. One of the  most visible fair abridgement cases in 1870s, Lord Hardwick, influential, said that the doctrine was part of the Statute of Anne.  Fair abridgement = create a new book. Not an excuse: if fair abridgement, then no infringement. Amount taken; effect on market; added as developed. Not modern fair use but close in some respects. Considered this to be a Q of fact, though difficult for a common jury. He preferred to adjudicate the issue in chancery, aided by experts appointed by parties, but didn’t rule no jury could hear it. Partisan expert testimony became more common in courts of law just as he was leaving. He specifically said that his job was to interpret the statute in equity the same way it would be interpreted at law.

Elephant in the room: no known fair abridgement cases at law before 1800. So what do we do? In 1791, nothing prevented a P from raising the issue in law and having a jury determine it, since it came from the statute, was a component of the infringement analysis, and had been called an issue of fact. Maybe the law courts would have rejected Lord Hardwick’s interpretation, but they didn’t reject it in 1807 when they got the chance.

Discussion period

Samuelson: Haiku? Popularity? It’s not popularity, but use of your learning as a programmer v. having to learn a new language or dialect in order to write for the platform.

Compatibility: there are programs in Java that run on Android and vice versa. It’s true they’re not fully interoperable but that’s not necessary. Oracle is arguing that you have to copy the whole thing for interoperability and that’s not a sound analysis. The 83 software engineers’ brief is a really good explanation.

Capella: New languages are common; programmers are used to switching. That’s where a lot of innovation can come from. APIs are a tool/part of the overall system: would be more relevant to look at APIs themselves. [which is why these are highly functional!]

Samuelson: Fed Cir rejects six other appellate decisions that say interfaces are uncopyrightable to the extent they facilitate/enable interoperability. Fed Cir didn’t just distinguish those decisions as being about “true” compatibility (and Connectix is about replacing the other guy’s platform and is only partly compatible)—it’s a rejection of every other decision except Jaslow.

Capella: There’s different types of interfaces. APIs are tools that programmer can use. Interfaces to platforms are very different types. Declaring code isn’t just for APIs, but even for internal modules that can be called. Slippery slope.

Brannen: A lot of cases interpret 102(b) to codify idea/expression. Is there any way to give effect to the words of the statute w/o having the exception swallow the rule. Worried about second class citizenship for software. 102(b): should you just ask yourself, did they copy expression that had merged?

Samuelson: It’s correct that the Fed Cir said you can’t take 102(b) completely literally b/c it would prevent all (c) for computer programs. But it’s not correct to say that all the words in that statutory provision but “idea” should be ignored. There is a clean distinction b/t things that are the method/system versus explanation/expression. Not everything creative in a work is protectable by (c) law.

Q: size Q: 11,500 lines/less than .5% of code in Java and .1% in

RT: look at the work registered. APIs are tools; the whole work, Java SE, is protectable but parts of it aren’t.

Brannen: Qualitatively: they could have written it differently? They didn’t b/c what they copied was popular w/the developer base they wanted it to court

RT: That’s not what Fed Cir said; they said it was qualitatively important b/c Google took it which was circular/inherently anti-fair use.

Actually, some things are free: 107 says so! It even says that “multiple copies for classroom use” are examples of fair use, not just criticism.

Popularity is one way to say it; utility is another. QWERTY keyboard isn’t popular because people like the expression; it’s popular b/c it makes transferring between keyboards easy. If I can read your shorthand b/c we both trained on the same system, that’s different from both of us liking Harry Potter.

Cappella: it was popular because it was elegant and people used it.

Samuelson: if you read the lower court opinion, you can see that what this was about was that Java was originally developed for enterprise systems. Smartphone was going to be a different kind of computing environment; needed new APIs, but for things like comparing two numbers and seeing which was larger, those things needed to be done on both environments and should be kept together.

Brannen: there was expression; troubling if they could pick what they wanted w/o paying and kill Java for mobile.

Q: does the SCt need to understand all this to resolve it? Google describes it as functional and Oracle as expressive.

Samuelson: that’s one reason many of us are nervous about this case. By comparison w/Lotus v Borland, this is more technically complicated.  Some of the big players really want the copyrightability issue decided so they’re doing the best they can to explain what’s at stake. Only Breyer has followed the issues; would expect him to try to take the lead on that.

Capella: Q of who should evaluate this—it’s important to look at it from the audience’s viewpoint, which is the app developers. Trying to evaluate expressivity in a foreign language, which is difficult.

RT: There was testimony that both the trial judge and the jury relied on about what professionals thought about protectability—which is one reason this case may be tempting for the Court to see as a standard of review case. This could be another Inwood v. Ives, which is an important TM case but at the Court is decided as standard of review. And the Court may be better at getting that right, given what it knows, than at evaluating the technical details of ©ability [or, as Jay suggests, at evaluating the history].

Brennan: does gov’t’s position matter?

Jay: Maybe, but has been more pro-© claimant in a number of cases than the Court has been. The fact the gov’t thinks it’s © eligible is not going to be institutionally persuasive.

Samuelson: SG’s position in Georgia v. PublicResource was rejected; doesn’t think it’s hugely impactful.

Q: standard of review/7th Amendment?

Gomez-Arogstegui: The appellate standard of review is de novo; that’s easy. The rule 50(b) standard the trial court is supposed to apply is also easy—it’s deferential. What’s tricky is that 50(b) assumes that an issue of fact has been properly assigned to the jury, which is in doubt given Harper & Row.  You can’t call fair use a mixed Q of fact and law if it turns out you have a constitutional right to all issues relating to fair use—would have to either overrule Harper & Row or distinguish it as involving a bench trial.

Snow: Harper & Row is the final twist in this unusual case. History until H&R rarely involved SJ in fair use cases; 9th Circuit starts to in the early 1980s, and H&R then produces Fisher v. Dees in which 9th Cir. says it’s pure law. 2d Circuit stays reluctant, but in 1990s everyone starts granting SJ, including Second Circuit in Castle Rock (Sotomayor). Key Q: Does the court want to go back to when fair use couldn’t be decided on SJ? May seem like lesser evil v. wading into the tech, but then all these cases have to go to trial. It’s a bit silly to have so much turn on 14 sentences from hundreds of years ago in one case. But it could be good to clarify that Harper & Row’s language was referring to a bench trial; jury verdicts post Harper & Row are still reviewed on a deferential standard.

Jay: review of jury verdicts is deferential, but as to what? If a pure legal Q surrounds why the issue goes to the jury, everyone agrees that in the JNOV decision the judge can reexamine the legal Q and decide it wasn’t legally sufficient to go to the jury. Might be able to avoid judge/jury issue if they resolve the law/fact issue in certain ways; not all mixed Qs are alike.

Snow: we can’t dodge the constitutional issue if the inferences from the facts have historically been sent to the jury and so history shows there’s a right to have the jury make the inferences.

Samuelson: most unsatisfactory about the fair use reasoning: it said there are only 2 facts we’re willing to recognize—whether there was good/bad faith and whether it was more functional than expression. If you look at Alsup’s opinion, there were a number of other factual issues where the jury was presented with evidence on the parties’ positions. If the SCt goes in the fair use direction, hopes they’ll broaden the set of issues considered “fact” issues. How you can possibly say market harm is a purely legal issue is a Q!

RT: slightly different take b/c there are areas of normative/legal concern where, for example, we just don’t consider the © owner’s claim that it was willing to license parodies as evidence of market harm. So for each factor there can be normative components and factual components.

Josh Sarnoff: hard to reach fair use w/o finding protectability of what was copied [and yet courts do it regularly by using thinness of © in the fair use factors; that’s treated an amicus I filed in the pending 2d Cir. Warhol case]. US Bank is fundamentally incoherent: the mixed Q is “does the law apply to these facts?” But giving the overall fair use Q to juries makes no sense at all. The Court could explain this properly, though it wouldn’t solve “what conduct is fair or not fair?” The Q of what conduct is illegal under the statute—that’s a judge question.

Snow: thinks it’s more like negligence/reasonability of conduct.

Sarnoff: thinks it’s more like obviousness in patent; among other things was always a statutory question and not a common law question.

Q re arguments

Jay: J. Thomas has historically written IP opinions but not asked questions; the new format allows him to ask questions that can be agenda-setting, while J. Breyer will have a time limit which is also new.

Q: there were 60 amicus briefs filed, many focused on innovation/software industry. We’ve seen Court punt on policy issues to Congress and say it was just applying the statute. Will the Court engage w/arguments about innovation?

Samuelson: thinks Court will care about industry structure, but it’s hard to get attention for one particular brief if there are so many. Clerks will try to identify a relatively small number of briefs for Justices to read. Microsoft/IBM briefs are likely to get attention, and computer science briefs on both sides. Comparison to eBay v. MercExchange, where innovation was something the Court took seriously.

Brennan: even if they don’t address it explicitly, hopefully they’re thinking about it given the outpouring of concern. Some of the companies on Google’s side have shifted to a nonproprietary model where they give software to people for “free,” but remember that nothing’s really free. Proprietary software still matters and we shouldn’t destroy that model.

Q: For Capella: how does she think about the patent/© interface for software, as a former software programmer? Patent trolls have been a problem, but © lasts so long.

A: Original expression should be protected; there wasn’t a lot of open source when she was practicing, but even open source involves tradeoffs depending on the license. You’re not getting something for free. If you use and build on innovation of others, many licenses have obligation to give back. Some companies have chosen to go that direction. Have to abide by the law.