Friday, October 9, 2020 [Tweets] [Favorites]

Google v. Oracle at Supreme Court

Timothy B. Lee:

The Supreme Court’s eight justices on Wednesday seemed skeptical of Google’s argument that application programming interfaces (APIs) are not protected by copyright law. The high court was hearing oral arguments in Google’s decade-long legal battle with Oracle. Oracle argues that Google infringed its copyright in the Java programming language when it re-implemented Java APIs for use by Android app developers.

John Gruber:

My gut feeling is that Google is in the right here — APIs should not be copyrightable — but that they utterly failed to make the argument in a clear way.

See also: Miguel de Icaza, Florian Mueller (3, 4).

Charles Duan:

Readers of this site no doubt know that Oracle’s arguments in its lawsuit against Google, set to be argued in the Supreme Court on Wednesday, could spell disaster for the computer industry, by turning the act of reimplementing an API into copyright infringement. Back in January, I revealed in an Ars Technica piece that it could even spell disaster for Oracle itself, because Oracle’s cloud storage service reimplements Amazon’s S3 API. Oracle did not dispute my findings but shrugged them off, claiming Amazon had granted permission. I was skeptical, but at the time did not have hard evidence to prove a negative that Oracle had no license.

I’ve now found the evidence for why Oracle should be worried. And more importantly, it shows why every tech company and startup should be worried about the Google v. Oracle case.



This is unfortunate, but not surprising. When IBM released its PC in the 1980s, its BIOS, the basic operating system that shipped on the ROM, was reverse engineered and cloned, but the cloners had two completely independent teams, one to reverse engineer and produce a specification and another to implement that specification. The goal was to obviate the copyright protection since no one could copyright an idea.

Google seems to have just copied the code.

In the long lasting fight over the UNIX API, there had been an obvious borrowing of ideas, but, as eventually decided by the courts, no copyright violation. I gather one sticking point was over a file describing ASCII characters which came down to arguing that there was really no way to express the idea without apparent similarity.

It is possible Google did have an argument to make. Perhaps they did go through some isolating process, but they sure didn't make that case in court.

Old Unix Geek

This is so insane it boggles the mind. Taken to its logical conclusion, everyone would need a license merely to compile a program to x86. After all, Intel and AMD invented that API. It's analogous to asking Oracle to pay each time it uses the English language. Furthermore, it's not as if Oracle invented the concept of many of the functions under discussion such as "max" -- they existed in mathematics previously.

It's really sad to see what Oracle has done to the corpse of Sun Microsystems. This is another example of rent-seeking by a corporation that can no longer grow the pie for everyone, therefore just seeks to grow its slice at everyone else's expense. We're seeing more and more of that, these days, unfortunately.

> Taken to its logical conclusion, everyone would need a license merely to compile a program to x86. After all, Intel and AMD invented that API.

No, you are conflating reimplementing and API and consuming it. You must buy a license to design processors that implements the x86 instruction set (just like you must have one to implements the ARM instruction set).

You don't need anything to write software that run on ARM or Intel processors.


No I don't think I am.

You're thinking of patents. Patents claim ownership over a particular way in which to implement functionality. Copyright claims ownership over the expression of an idea.

Cyrix didn't use Intel's patents, and did a clean room implementation: they implemented an x86 chip without using any of Intel's patents, only its "APIs" (the instruction formats and the data formats used for paging for instance). Intel doesn't license its instruction set to third parties. Compaq did a clean room implementation of IBM's BIOS in the same way. Those have been legal because APIs were not copyrighted.

The same arguments hold for all programs using those APIs. If you allow the non-copyrighted lingua-franca between libraries/CPUs and software to be copyrighted, every single header file you include into your program, every single use of an x86 instruction will be a use of copyrighted information: you are using the expression of an idea. There's no reason to encode the "return" instruction by the byte 0xC3: that's expression not functionality. This is how insane a mess this ruling could become.

This lawsuit, if Oracle wins, will be so damaging to the whole world (not just the computer industry) that it boggles the mind. Is there anything we (the tech community as a whole) can do to rein in, and punish, Oracle for this evil?

In the old, old days, Sun Microsystems was a client of mine. The people I knew there seemed nice, but when Oracle took over it turned into crap. The insider joke was that "Oracle" was really an acronym, standing for One Real Asshole Called Larry Ellison. Seems nothing much has changed, sadly.

Stay up-to-date by subscribing to the Comments RSS Feed for this post.

Leave a Comment