Bradley M. Kuhn

• Contact
• Blog  
• Interviews / Articles  
• Software
• Résumé

Considerations For FLOSS Hackers About Oracle vs. Google

Monday 16 August 2010 by Bradley M. Kuhn

Many have already opined about the Oracle v. Google lawsuit filed last week. As you might expect, I'm not that worried about what company sues what company for some heap of cash; those sort of for-profit wranglings just aren't what concerns me. Rather, I'm focused on what this event means for the future of software freedom. And, I think even at this early stage of the lawsuit, there are already a few lessons for the Free Software community to learn.

Avoid Single-Company-Controlled Language Infrastructure

Fourteen months ago, before the Oracle purchase of Sun, I wrote about the specific danger of language infrastructure developed by a single for-profit patent-holding entity (when such infrastructure is less than 20 years old). In that blog post, I wrote:

[Some] might argue that with all those patents consolidated [in a single company], patent trolls will have a tough time acquiring patents and attacking FaiF implementations. However, while this can sometimes be temporarily true, one cannot rely on this safety. Java, for example, is in a precarious situation now. Oracle is not a friend to Free Software, and soon will hold all Sun's Java patents — a looming threat to FaiF Java implementations … [A]n Oracle attack on FaiF Java is a possibility.

I'm sorry that I was right about this, but we should now finally learn the lesson: languages like Java and C# are dangerous. Single companies developed them, and there are live, unexpired patents that can easily be used in a group to attack FaiF implementations. Of course, that doesn't mean other language infrastructures are completely safe from patents, but I believe there is greater relative risk of a system with patent consolidation at a single company.

It also bears repeating the point I made on Linux Outlaws last July: this doesn't mean the Free Software community shouldn't have FaiF implementations of all languages. In fact, we absolutely should, because we do want developers who are familiar with those languages to bring their software over to GNU/Linux and other Free Software systems.

However, this lawsuit proves that choosing some languages for newly written Free Software is dangerous and should be avoided, especially when there are safer choices like C, C++, Python, and Perl⁰. (See my blog post from last year for more on this subject.)

Never Let Your Company File for Patents on Your Work

James Gosling is usually pretty cryptic in his non-technical writing, but I think if you read carefully, it seems to me that Gosling regrets that Oracle now holds his patents on Java. I know developers get nice bonuses if they let their company apply for patents on their work. I also know there's pressure in most large companies to get more patents. We, as developers, must simply refuse this. We invent this stuff, not the suits and the lawyers who want to exploit our work for larger and larger profits. As a community of developers and computer scientists, we must simply refuse to ever let someone patent our work. In a phrase: just say no.

Even if you like your company today, you never know who will own those software patents later. I'm sure James Gosling originally never considered the idea that a company as revolting as Oracle would have control of everything he's invented for the last two decades. But they do, and there's nothing Gosling can do about what's done with his work and “inventions”. Learn from this example; don't let your company patent your work. Instead, publish online to establish prior art as quickly as possible.

Google Is Not Merely a Pure Free Software Distributor

Google has worked hard to cast themselves as innocent, Free-Software-producing victims. That's good PR because it's true, but it's also not telling the whole truth. Google worked hard to make sure Android was completely Apache-2.0 (or even more permissively) licensed (except for Linux, of course). There was already plenty Java stuff available under the GPL that Google could have used. Sadly, Google was so allergic to GPL for Android/Linux that they even avoided LGPL'd components like uClibc and glibc (in favor of their own permissively-licensed C library based on a BSD version).

Google's reason for permissive-only licensing for “everything but the kernel” was likely a classic “adoption is more important than software freedom” scenario. Google wants Android/Linux in as many phones as possible, and wants to eliminate any “barrier” to such adoption, even if such a “barrier” would defend software freedom.

This new lawsuit would be much more interesting if Google had chosen GPL and/or LGPL for Android. In fact, if I fantasize about being empowered to design a binding, non-financial settlement to the lawsuit, the first item on my list would be a relicense of all future Android/Linux systems under GPL and/or LGPL. (Basically, Google would license only enough under LGPL to allow proprietary applications, and license all the rest as GPL, thus yielding the same licensing consequences as GNU/Linux and GNOME). Then, I'd have Oracle explicitly license all its patents under GPL and/or LGPL compatible licenses that would permit Android/Linux to continue unencumbered, but under copyleft. (BTW, Mark Wielaard has a blog post that discussed more about the issue of GPL'd/LGPL'd Java implementations and how they relate to this lawsuit.)

I realize that's never going to happen, but it's an interesting thought experiment. I am of course opposed to software patents, and I certainly oppose companies like Oracle that produce almost all proprietary software. However, I can at least understand the logic of Oracle not wanting its software patents exercised in proprietary software. I think a trade off, whereby all software patents are licensed freely and royalty-free only for use in copylefted software is a reasonable compromise. OTOH, knowing Oracle, they could easily have plans to attack copyleft implementations too. Thus, we must assume they won't accept this reasonable compromise of “royalty-free licensing for copyleft only”. That brings me to my next point of FaiF hackers' concern about this lawsuit.

Never Trust a Mere Patent Promise; Demand Real Patent Licenses

I wrote after Bilski that patent promises just aren't enough, and this lawsuit is an example of why. I presume that Oracle's lawyers have looked carefully as the various promises and assurances that Sun made about its Java patents and have concluded Oracle has good arguments for why those promises don't apply to Android. I have no idea what those arguments are, but rarely do lawyers file a lawsuit without very good arguments already prepared. I hope Oracle's lawyers' arguments are wrong and they lose. But, the fact that Oracle even has a credible argument that Android/Linux doesn't already have a patent license shows again that patent promises are just not enough.

Miguel de Icaza used this opportunity to point out how the Microsoft C# promises are “better” by comparison, in his opinion. But, Brett Smith at FSF already found huge holes in those Microsoft promises that haven't been fixed. In fact, any company making these promises always tries to hide as much nasty stuff as it can, to convince the users that they are safe from patent aggression when they really aren't. That's why the Free Software community must demand simple, clear, and permanent royalty-free patent licenses for all patents any company might hold. We should accept nothing less. As mentioned above, those licenses could perhaps require that a certain Free Software copyright license, such as GPLv3-or-later, be used for any software that gets the advantage of the license. (i.e., I can certainly understand if companies don't want to accidentally grant such patent licenses to their proprietary software competitors).

Indeed, it's particularly important that the licenses cover all patents and those possibly exercised in future improvements in the software. This lawsuit has clearly shown that even if patent pools exist for some subsets of patents for some subsets of Free Software, patent holders will either use other patents for aggression, or they'll assert patents in the patent pools against Free Software that's not part of the pool. In essence, we must assume that any for-profit company will become a patent troll eventually (they always do), and therefore any cross-licensing pools that don't include every patent possible for any possible Free Software will always be inadequate. So, the answer is simple: trust no software-patent-holding company unless they give an explicit GPLv3-compatible license for all their patents.

We Must End Software Patents

The failure of the Bilski case to end software patents in the USA means much work lies ahead to end software patents. The End Software Patents Wiki has some good stuff about this case as well as lots of other information related to software patents. There are now heavily funded for-profit corporate efforts that seek to convince the Free Software community that patent reform is enough. But, it's not! For example, if you see presenters at FLOSS conferences claiming to have solutions to patent problems, ask them if their organization opposes all software patents, and ask them if their funders license all their patents freely for GPLv3-or-later software implementations. If you hear the wrong answers, then their motives and mission are suspect.

Finally, I'd like to note that, in some sense, these patent battles help Free Software, because it may actually teach companies that the expense of having software patents is not worth the risk of patent lawsuits. It's possible we've reached a moment in history where it'd be better if the Software Patent Cold War becomes a full Software Patent Nuclear War. Software freedom can survive that “nuclear winter”. I sometimes think that in the Free Software community, we may find ourselves left with just two choices: fifty more years of Patent Cold War (with lots of skirmishes like this one), or ten years of full-on patent war (after which companies would beg Congress to end software patents). Both outcomes are horrible until they're resolved, but the latter would reach resolution quicker. I often wonder which one is the better long term for software freedom.

But, no matter what happens next, the necessary position is: all software patents are bad for software freedom. Any entity that supports anything short of full abolition of software patents is working against software freedom.

---

⁰I originally had PHP listed here, but jwildeboer argued that Zend Technologies, Ltd. might be a problem for PHP in the same way Oracle is for Java and Microsoft for C#. It's true that Zend is a software patent holder and was involved in the development of later PHP versions. I don't think the single-company-controlled software patent risks with PHP are akin to those of Java and C#, since Zend Technologies isn't the only entity involved in PHP's development, but certainly the other languages listed are likely preferable to PHP.

Posted on Monday 16 August 2010 at 07:20 by Bradley M. Kuhn.

Comment on this post in this identi.ca conversation.