Declaring Code Is Not Code, Says Larry Page (arstechnica.com)

Alphabet CEO Larry Page says his company never considered getting permission from Oracle for using the latter's Java APIs in Android. Page, who appeared in a federal court, said Java APIs are open and free, which warrants them or anyone to use it without explicit permission from Oracle. From an Ars Technica report (edited for clarity): "But you did copy the code and copy the structure, sequence, and organization of the APIs?" Oracle attorney Peter Bicks asked, raising his voice. "I don't agree with 'copy code,'" Page said. "For me, declaring code is not code," Page said. "Have you paid anything to Oracle for using that intellectual property?" Bicks asked. "When Sun established Java, they established it as an open source thing," Page said. "I believe the APIs we used were pretty open. No, we didn't pay for the free and open things." [...] "Was Google seeking a license for Java?" Google lawyer Robert Van Nest asked. "Yes, and a broader deal around other things, like branding and cooperation," Page said. "After discussions with Sun broke off, did you believe Google needed a license for APIs?" Van Nest asked. "No, I did not believe that," Page said. "It was established industry practice that the API and just the headers of those things could be taken and re-implemented. [It must be done] very carefully, not to use any existing implementation of those systems. That's been done many, many times. I think we acted responsibly and carefully around these intellectual property issues."

Comments

Giant problem

[NotInHere]

If APIs are copyrightable, this will be a huge problem for projects like Wine (which implements Microsoft APIs), and GNU/Linux (which implements Bell labs APIs).

Re:Giant problem

[LichtSpektren]

And also Microsoft. Remember that MS-DOS was a renaming of Seattle Computer Products' QDOS, which itself was an API clone of CP/M.

If the jury rules for Oracle, that means Microsoft will owe billions to the estate of Gary Kildall.

Re:Giant problem

[segedunum]

Indeed so. Software is in a whole heap of trouble of APIs are copyrighted and you have to go to court to prove whatever laughable versions of fair use is in vogue this week.

Re:Giant problem

[StormReaver]

If the jury rules for Oracle, that means Microsoft will owe billions to the estate of Gary Kildall.

And IBM will be able to sue Oracle and Microsoft for billions, for Oracle's use of the SQL API .

Re:Giant problem

[peragrin]

Only after SCO sues IBM for trillions.

Oracles argument was used in SCO vs IBM and it was tossed there too. Headers are not copyright able.

Re:Giant problem

[cyriustek]

Organisations publish their APIs, because that want people to use them.

Sun was making a huge push on this in the early 2000s. IMO Larry Page is spot on with this.

Re:Giant problem

[Solandri]

Not really. Software companies will all close shop in the U.S. and move their operations to countries where APIs are legally declared not copyrightable. All those companies hiring H1B programmers from India or outsourcing programming work to India? They'll move to India and if you're lucky they'll outsource some of their work to you in the U.S. Software development will continue on in the rest of the world as if nothing had happened. The U.S. will be relegated to a software backwater, as most of the software made and sold in the rest of the world cannot legally be distributed in the U.S.

That's the nature of the free market. It interprets stupidity as damage, and routes around it.

Re:Giant problem

[I4ko]

Next thing you know - UK suing US for copyright infringement on English language since US stopped paying their license fee with the Boston Tea Party.

Has Sun/Oracle ever copied any APIs?

[jonwil]

Someone should go back and look for any examples where Sun or Oracle have copied APIs but didn't have any specific license to use the code behind those APIs.

There must be some example somewhere of Sun or Oracle doing exactly what they are now claiming Google has done...

Re:Has Sun/Oracle ever copied any APIs?

[jfdavis668]

Isn't Java based on C syntax? You think they would have copied the API somewhere to do that.

The API _is_ the semantics of language

[mysidia]

An Ad-hoc one, but part of language nonetheless.

The ENGLISH Language has an API too. You will find much of it documented in a dictionary.

The words are the element of the language; but an API tells you how to exchange messages between two people.

Attempting to have exclusive rights to an API is like a restaurant wanting exclusive rights to phrases such as "GET WATER", or "ONE BEER PLEASE".

So patrons will be sued if they go to a competitors' restaurant and formulate requests such as that

The code does things...... the API is just a functional (non-creative) description of the correct way to interact with the code.

If Oracle wins, get into corporate law

[LichtSpektren]

If Oracle wins this case and all the appeals, where it's ruled that using open APIs is copyright infringement, then I would strongly suggest you get into corporate law.

Because the end result is that basically every software company in the world (including not just Oracle and Google but Microsoft, Apple, IBM, Intel, Samsung, etc.) will suddenly find themselves in a Mexican standoff of potentially trillions of dollars in "intellectual property" suits ready to be fired off. The only winner will be the one with the best legal department; oh, and the lawyers.

All the aforementioned companies would be wise to pen amicus curiae letters in favor of Google.

Re:Compaq copied APIs from IBM

[drerwk]

Compaq reverse engineered the BIOS. It did not copy the text of some file that defined the API. I don't think that the INT operations even had fixed names - they had numbers. So a BIOS call would be documented as http://stanislavs.org/helppc/i...
INT 16,0 Wait for keystroke and read
This exact operation can be described using different words e.g.
On Int 16,0 the system will pause until a keystroke is pressed and the value will be placed in AH.
I understand the issue in Oracle v. Google to be exact copying of some number of interfere files. Such files did not exist for BIOS as far as I recall.

Bloody IP

[SlashDread]

Riddle me this.

Intellectual means "of the intellect" and is thus intangible.
Property has always been used as a nomer for physical items that are clearly in possession, after all possession is 9/10th of the law.

This whole "IP" terminology is thus clearly double speak, and should be avoided. The whole legal constructs around them, be it patents, invention or copyright are only there to not disrupt existing economic structures. They are a philosophical abomination, especially in the digital age, where copying is cost-less, and distribution nearly so. This is true for books, code, movies and basically everything digital IMHO, and in this case even more so.

In this particular case of Oracle vs Google regarding Java "IP" we are talking about API Headers. To anyone with some coding background, API Headers are a description of a system. They are not patentable, as patents require implementation. (In Europe software is considered "math", and atm not patentable at al) They should not be copyrightable, for the same reason that announcing you will write a book about Fire and Ice and Dragons is a description of a book, but not the book itself. This description should not grant you the right to be the sole author of books about Fire and Ice and Dragons.

I applaud Google in this fight, and I hope they fight till they win.

Books, Music, and APIs

[rockmuelle]

Ok, I'm going to take a slightly unpopular stance here and suggest that APIs probably should be copyrightable.

Ignoring all the legal issues, my rational is simple: An API spec represents the output of the intellectual effort of the architect far better than any implementation code. Designing a good API is difficult. Doing so requires finding exactly the right abstractions to allow users to use your API to perform complex operations in a simple, straightforward manner. The design process often involves several iterations of implementation to refine the API - not only implementing the functionality, but writing software that uses to to make sure it meets its goals.

The book title/chapter title argument is often used to show why APIs shouldn't be copyrighted. That's a poor analogy. Many books don't bother with chapter titles, so they're clearly not essential to the interpretation of the material. When present, they usually can't be used to quickly summarize a book. A well designed API will clearly and concisely present to you everything the underlying library can do. In fact, you should never care about the implementation of the library. If you ignored the text of a book, you wouldn't really be reading it.

If we're going to look to artistic pursuits for analogies, I'll suggest music is a better one. Melodies and lyrics are primarily what's copyrightable in music. Chord changes and musical embellishments are not. Arrangements are copyrightable when written down. Specific performances are also copyrightable (which is why samples must be cleared for use in other songs). Melodies and lyrics are akin to an API - the instantly let you identify the song/library and are the primary way most people remember it. For example, I can play "Yesterday" on a piano, guitar, speak-and-spell, and it's still a Beatles song. I still owe the Beatles royalties for using their lyrics and melody, regardless of how I arrange and perform it.

Now, if we allow APIs to be copyrighted, we gain a lot of flexibility. Most importantly, the copyright holder can release the API under a free and open license if they want. The designer can say: here's my work, feel free to do with it what you want. Or, they can lock it down and restrict what can be done with it.

For better or worse, Sun wanted the best of both worlds with Java. They implied that it was free and open, but never actually released the APIs under a specific free and open license. In the music world, the "Happy Birthday" saga is similar. The melody and lyrics were thought to be under copyright and Warner collected a few million a year from artists for performing it and using it in their works. Just like the core Java APIs, everyone knows "Happy Birthday" and most people used it casually without paying royalties (yes, I realize not everyone knows Java, but most readers of /. do). With the song, it turned out an earlier version that wasn't under copyright was found, which invalided Warner's claims. Unfortunately, that's unlikely to happen with Java.

tl;dr: APIs are the creative output of the design process, just like melodies and lyrics are in music. They probably should be copyrightable.

-Chris

Re:Books, Music, and APIs

[PRMan]

Which MIGHT be true, except Sun said they were open source and free to use. Calling backsies was garbage on the playground and it's still garbage.

Re:Books, Music, and APIs

[dgatwood]

In general, the fact that something is a creative work or takes time to create doesn't necessarily make it copyrightable. For a creative work, any functional aspects of that design are supposed to be protected by patents, not copyright. And an insufficiently creative work isn't protected at all, no matter how much time it took to create it.

For example, the courts long ago ruled in Feist Publications, Inc., v. Rural Telephone Service Co. that the difficulty of creating something is not sufficient to make it copyrightable when they declared that a phone book is a non-creative collection of facts. One could reasonably argue that a header file collects the declarations from source code, and that the real creative work is the source code itself. After all, the sole reason for a header file is to consolidate a bunch of declarations into a form that that makes it easier for a compiler to digest. This arguably makes a header no more a creative work than the phone book. That's not arguing that an API shouldn't be copyrightable per se, so much as that a header shouldn't be unless it contains other creative works beyond the declarations.

Also, per 17 U.S.C. section 1302, anything that is "dictated solely by a utilitarian function of the article that embodies it" is not eligible for copyright. The intent of copyright law is to shift responsibility for protecting such creative works into the domain of patent law. So given that there's a strong utilitarian aspect to APIs (because any function has basically exactly one valid declaration, or else your code won't link correctly), if you want to argue that an API should be protected by copyright, you have to come up with a concrete argument of why that API's design is more than just utilitarian in nature. So any creative effort that was focused on making an API easier to use doesn't count towards the creativity requirement for copyright purposes.

Re:Books, Music, and APIs

[The Raven]

Ok, I'm going to take a slightly unpopular stance here and suggest that APIs probably should be copyrightable.

Ignoring all the legal issues, my rational is simple: An API spec represents the output of the intellectual effort of the architect far better than any implementation code. Designing a good API is difficult.

You forget that an important part of copyright is 'To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.' APIs are so fundamental to all creation, that being able to own an API would completely lock down and prevent the progress of science and useful art.

You can copyright a painting, but not a painting style. Even if your style of painting is new, innovative, unique, and you spent years developing your unique method... you do not get any control over others copying your technique and using it in their own works. (e.g. Picasso and Cubism, Seurat and Pointilism)

You can copyright a book, even a paragraph, but you cannot copyright a unique way of looking at the world. You could spend months on your ideas, on your unique take on a topic... but that does not grant you a copyright to the idea, only to the specific implementation of your paragraphs, chapters, and novels. (e.g. Tolkein and Elves, Niven and Ringworlds)

The reason that ideas (and, by extension, APIs) are not copyrightable is because the only way to claim ownership an idea is via Patents. Now you can have endless debates on what should or should not be patentable, how unique it is, and the merits (or lack) of software patents, but the end point is that if you believe your software idea deserves protection the only way you can is via a patent. Because copyright only protects a specific implementation of an idea, not the idea itself no matter how much work went into generating that idea.

Re:Type systems

[segedunum]

No it hasn't, throughout the decades copyright of software has been upheld.

Yes it has. Software implementations, yes, APIs no. Not sure if you're just ignorant or are deliberately painting over that. The developer tools market is built on this and has been since the year dot.

Nah, because there is a strong interoperability defense, that allows you to copy copyrighted APIs (which is what your friend dr dos did).

Nope, because every company is now going to have to go to court to prove 'fair use'', which Oracle has made various versions of as it has gone along. If you can't prove that then there are billions in free money waiting for everyone who can claim copyright over APIs. At the time of DR DOS APIs were certainly not copyrightable and haven't been deemed to be so until recently. There is no *strong* fair use defence. That was shoehorned in when the people making the laughable ruling realised what trouble it would cause, but it creates a nice walled garden nonetheless.

1) They didn't make a clean-room copy.

Yes they did, and it's not a copy. Dalvik isn't a Java.

2) They didn't make their copy for interoperability reasons.

Yes they did. They wanted a familiar language where developers could reuse *their* code and port over. There are developer tools that have been on the market for decades that will do exactly that. Oracle's idea that because Android isn't 100% compatible with Java, and they didn't copy *all* the APIs, so therefore it can't be about interoperability is absolutely laughable.

3) They didn't use any of the licenses that were available for Java. Google has now switched to the GPL for their Java, so there is no more problem for them going forward.

They didn't need a license because they didn't copy anything from Java and what they implemented isn't Java. What code you could write for Android looks like Java, as a language, but that's all.

It's a distinction Oracle are rather desperate to muddy because they'd like to make some money out of this case because they failed to develop anything remotely useful people wanted. A lot of their customers are starting jump ship as well. Revenue-wise this is quite important to Oracle, and you see that in how desperate their lawyers get.

Much like what already happened with encryption.

[Ungrounded Lightning]

Software companies will all close shop in the U.S. and move their operations to countries where APIs are legally declared not copyrightable. ... The U.S. will be relegated to a software backwater, as most of the software made and sold in the rest of the world cannot legally be distributed in the U.S.

Something similar to this happened with encryption. The US regulated it as a weapon and banned / limited / added red tape to the export of strong encryption software. US companies also couldn't import strong encryption software, include it in their products, and resell them elsewhere. The software had to be installed outside the US by non-US companies.

The result was that commercial development and deployment of strong encryption software pretty much stopped in the US and picked up outside its boundaries for several years, and various workarounds were developed (such as "encryption with a hole" so a strong encryption module could be installed later).

This continued up to about the turn of the centur, when laws, policies, and court decisions loosened things up enough that US companies could play again.