Solutions to the Software Patent Problem

Speech given at the Locatelli Center, Santa Clara University, in November 2012  (video,  metadata)

Andrew Chen: Thank you, Eric.

My name is Andrew Chen. I teach patent law at the University of North Carolina, and I had a previous life as a Computer Science professor.

I have the easiest job today, which is introducing two men who need no introduction. Richard Stallman, we know, is the founder of the free software movement, co-founder of the League for Programming Freedom, lead software architect for the GNU Project and author of Emacs, which he's described as a text editor and also a way of life. Something that I can agree with, having written my doctoral dissertation using his program.

Dr Stallman has decided not to participate in the live streaming facility for today. He explains that use of the streaming online would require use of the Microsoft Silverlight plugin, which would pressure people to use proprietary software. Dr Stallman considers it wrong to pressure people to do that. He would like you to know that he plans to make a recording of his presentation available at a later time in either the Ogg Theora or WebM formats.

Dr Stallman.


Richard Stallman: Can the tech people please confirm that the streaming is off?

OK, I think that's confirmation.

So, why are software patents bad? Or, “computational idea patents” as I think we should really call them, because each one is a monopoly on a computational idea. Most people, when you say “software patents,” they think it's a question of patenting a specific program. I'm sure all of you know that that's not what those patents do, but most people don't know that, so, to try to avoid misleading people, I call them “computational idea patents.”

So, anyway, the reason these are bad is that they deny people the freedom to use their computers as they wish and do their computing as they wish, freedom that everyone must have. These patents put all software developers in danger, and their users as well. A danger that there is no reason we should stand for. So: we should protect software from patents. Software needs patent protection: protection from patents.

But most people don't know enough about what patents do to appreciate why patents that can restrict software are so harmful. Most people think that patents are like copyrights, which is not true at all. The sum total of what they have in common is one sentence in the Constitution, and that similarity is so little and abstract it has nothing to do with the practical effects.

So, the last thing we should ever do is use the term “intellectual property” that confuses not just these two laws, but a bunch of other unrelated, disparate laws, that don't even share one sentence in the Constitution with those two. So that term spreads confusion whenever it's used and about eight years ago I decided I should never use it and I have never used it since then. It's surprisingly easy to avoid, because in general there's no reason whatsoever to use it except that it's chic. And once you learn to resist that, it's as easy as pie, just talk about one law, and then you call that law by its name, and you're making a coherent, clear statement.

So, I have to explain to people what patents do, and show them that it's not at all like what copyrights do. An analogy is a good way to do this. What can you say about programs? Well, they're large works, full of details that all have to work together to get the desired result. Well, what else is there like that? A novel, or a symphony. So, imagine if the governments of Europe in the 1700s had had the cockeyed idea of promoting the progress of symphonic music with a system of “musical idea patents.” So, any musical idea statable in words could have been patented. A melodic motif could have been patented, or a series of chords, or a rhythmic pattern, or a pattern of repetitions in a movement, or the use of certain instruments while the rest of the orchestra is silent and a bunch of other musical ideas that I can't think of, but maybe a composer would.

So, now imagine it's 1800 and you're Beethoven and you want to write a symphony. You're going to find that it's harder to write a symphony that you don't get sued for than to write a good symphony. Now, you'd probably have complained, and the patent holders would have said “Oh, Beethoven, you're just jealous because we had these ideas before you. Why don't you go think of some ideas of your own?” Of course, Beethoven is considered a great composer because he had lots of new ideas, and not only that, he knew how to use them effectively. And that is: combined with lots of familiar ideas, so that his pieces were merely shocking for a while, but people could get used to them. They were not so alien and incomprehensible that they got rejected. They shocked people for a while, people got used to them, and now we don't see what's shocking any more, because we're used to those ideas. Well, that's the proof that he used those ideas well.

So, the idea that anyone could, or should have to, reinvent music from zero, is absurd. Not even a Beethoven could do that, and it would be silly to ask someone to try. It's the same with computing. Just as a symphony implements many musical ideas together, but the hard part is not picking a bunch of ideas. The hard part is implementing them together with notes. It's the same with software. A large program will implement thousands of ideas together. But the hard part is not picking some ideas. It's easy to pick some ideas. What's hard is to implement them all together and make it work well.

So “computational idea patents” obstruct the hard and big job by promoting resources that we get plenty of anyways. So it's a misconceived system. Designed to give us help we don't want at the cost of tremendous problems.

So what we need is to get rid of the problem. What is the problem? The problem is: software developers and their users are threatened by patents. They are in danger. How can you prevent that? Well, one way is: don't issue patents that could affect software. That solution works if you apply it from the beginning. If a country never issues such patents, then its patent system doesn't attack software. OK, it's a good solution. But it's not applicable if a country has already issued hundreds of thousands of software patents.

Now, I've proposed that constitutions should explicitly say that patent privileges can be reduced just as they can be increased. That they are not in any sense somebody's property; they are privileges given out by the government which can be changed at will. After all, if you allow the government by legislation to increase them, it's absurd to make this a one-way ratchet. But that's not in the US Constitution.

So, what can we do? Well, we can ask courts to rule that all those patents that restrict software were invalid from the beginning and always have been invalid, and that gets rid of them all. However, that's not something that people can lobby for. It's not something we can say to officials, “do this because we want you to.”

So, if we're going to look for a solution that we can get implemented, what is there? Well, the only way I can see is to legislate that software is a safe harbor. If it's software, then you're safe. Circuits to do the same computation would be covered by a patent, but if it's software, then you're safe. But what does that mean? What does it mean for something to be software? Well, it runs on a general purpose, universal machine. So first you make a universal machine and then you put in the program to say what it should do. Well, if the machine's only function is to be universal, then the program is all that implements any specific, patented idea.

So, that's the case I want to get at, and I'm trying to separate it from a case like that in Diamond v. Diehr where there was a patent for a system, a method of curing rubber. The implementation involved a computer, but it also involved special purpose hardware, not a general purpose universal machine, and that special purpose hardware was crucial to carrying out the patented technique. It wasn't actually a software technique. And, actually, I read an article by Pamela Samuelson arguing that the CAFC twisted that decision and basically got the quantifiers in the wrong order. That the Supreme Court said, “the fact that there's a computer in there somewhere doesn't automatically make it non-patentable,” and the CAFC twisted that into “the computer makes it patentable.”

Anyway, we might have some hope with the courts, but I'm proposing a method that will separate the cases that we must protect from non-computational idea patents that affect systems that might be implemented with a computer in there somewhere. The precise words to use? Well, the best I could come up with was: “software running on generally used computing hardware.” We certainly want things like smartphones to be covered; we don't want it to exclude anything that has any kind of special-purpose hardware in there. The portable phone obviously has specialised hardware to talk to the phone network, but that shouldn't automatically mean that if it's running on a portable phone, it's vulnerable to patents. Because that is a general purpose computer and people use it for all sorts of things. But my words, “generally used computing hardware,” they may not be the best possible words. This is a subject that I think calls for study, because we've got to look at each possible wording that might be used and see which cases would be protected from patents and which would be exposed to come up with the right method.

Now, every time I suggest a method to solve this problem, the first thing people try to look for is how to half solve it instead. The idea of really solving the problem shocks people because it strikes them as radical. They think “I can't advocate something so radical as to really solve this whole problem. I've got to look for some partial solution that will only protect some software developers.” Well, that's a mistake. It's a mistake a) because it wouldn't do the whole job, but b) because it would be harder to get it passed. There are lots of software developers and they are all threatened and if we propose to protect them all, they will all have a reason to support it. But if we propose to only protect some of them, the rest will say “well, this doesn't do me any good, why should I care?”

So, let's propose a real solution. And, besides, partial solutions tend to be vulnerable to the problem that Boldrin and Levine have written about very effectively, that it's easy for the pressures for patents to stretch the boundaries if you give them any kind of boundary that they can stretch. And this, by the way, is another advantage of applying a change to suing people, rather than to what's patentable. Because there, the criteria are just “what kind of situation is this?” It's harder to stretch those, and if they tried, it would always be in a case against somebody who's going to be fighting not to stretch it. So it's less vulnerable to being distorted from an intended restriction of substance into an actual requirement of form of patent applications, which tends to happen to any kind of requirement about what patent applications have to look like.

So, there I go.


Andrew Chen: Thank you, Dr Stallman.