Thu, 14 Oct 2004

The trouble with software patents

Groklaw is running a piece on software patents, from someone who has a software patent and thinks we risk "throwing out the baby with the bathwater" if software patents are not available.

The question is not "do I have to right to get a patent on my idea?" but "is progress better served by granting software patents or not?". There is one school of thought that says if US patent system flaws were fixed, software patents would be OK. Very few people with experience doubt that the serious flaws in the US patent system could be fixed, and note the flaws in other countries heading in the same direction.

For proof of this, it was recently estimated that Linux violates 283 patents. Linux is mainly a reimplementation of Unix, a system which predated software patents by a long way: these patents are therefore "incidental" violations committed in the normal process of implementing software. This does not bode well for production of software: you will violate patents, you won't know it, and your work is no longer yours to control.

But let us assume, for a moment, that the levels of obviousness and prior art are dramatically increased, and all these patents are swept away. I can't see how such a thing is possible, but that's another debate.

We are still left with the non-obvious, novel patents, such as in the RProxy patent woes. Do we need patents as an incentive for people to make progress in software? How much damage are patents doing to software progress?

For the first question, the flourishing of software prior to its patentability shows this clearly, such as the case of Dan Bricklan's Visicalc (the first spreadsheet, not patented, now a market dominated by Microsoft's Excel). This is because, unlike drugs or mechanical devices, software is already protected by copyright: the extension of patents to cover software made it the only area covered by both. Indeed, copyright on software is far more powerful than on a book, because the copyright holder can distribute only the binaries, not their actual creation, leveraging into a monopoly on support and fixes. With an estimated 10 million programmers in the world, a lack of incentives is very hard to argue: this is a result of the barriers to entry being so low. This low barrier also makes things like Open Source development possible, and the Internet. It is hard to argue that we would have been better off without these developments.

On the damage side of the equation, these same low barriers to entry make the barriers produced by patents disproportionately destructive. If it costs millions to produce a drug, patent and lawyers fees don't make such a difference. It is the normal low barriers of software which make ubiquitous infrastructure possible. Once again, the rproxy problems illustrate the loss we face due to patents in this area.

[/IP] permanent link