Software Patents

There’s been plenty of digital ink spilled over the patent system in the wake of Google’s shameless hypocrisy on the matter. The conversation seems to move pretty fluidly between discussion of patents in general and discussion of software patents. My understanding is that the portfolio Google is complaining about involved more than just software patents, but I may be wrong; the point is that arguments about software patents is at least partially orthogonal to the Google situation.

Like the vast majority of software engineers I know, I’d much prefer that software were not patentable. But my main complaint with those defending the status quo is over arguments like this (quoted here):

Nobody has ever told me exactly why the patent system is fundamentally broken.

Whatever the underlying point, this is a textbook example of framing the discussion such that major catastrophic problems are necessary to overturn a product of history. In my experience, “fundamental” is an extremely ambiguous term, particularly where brokenness is concerned. By some definitions, any problem that can be hidden, at least temporarily, does not count as “fundamental”. Arguing that something is broken can degenerate into a whack-a-mole game, where you point out a negative consequence of some issue you consider fundamental, a way to hide that particular manifestation is devised, another similar negative consequence of the same issue is discovered, and another patch devised, and so on. Finding bugs is almost always harder than fixing them, so there’s a serious asymmetry of effort here. If the rules are that we play until someone is exhausted, the defender will inevitably win. After all, the defender is happy to invest their time in a system they want and expect to continue on to the future; if the attacker wins the argument and the “fundamentally broken” system is scrapped, then they get no benefit from all the investment they’ve made understanding the system. This situation arises in everything from terrible software architecture to ad-hoc reasoning frameworks.

But as for software patents, the focus of the argument shouldn’t be on whether a system for software patents could be fixed; it should be on whether there is any value to software patents at all. The patent system was created as a way to foster innovation, at the price of a well-understood evil: monopoly power. There’s never been any debate that patents have negative consequences. Thus it’s worth asking what the positives are.

I can’t speak to drug patents, or manufacturing patents, or wireless technology patents. Maybe in those fields a big chunk of innovation is performed for the purpose of getting patents. But I’d argue that there is more innovation in software than in every other field put together—new algorithms, human interfaces, programming languages, and engineering methods are being experimented with constantly. And 99.9% of all this innovation never gets patented in any way, shape or form. I don’t mean that 99.9% isn’t patentable; I’m saying that 99.9% of software that probably could be patented under the current system—precisely the types of innovation the patent system was created to reward—isn’t patented. Thus the absolute upper limit on what we’d lose without software patents is 0.1% of innovation.

But I’ll go farther than that. I have never encountered a single piece of software that was created primarily for the purpose of claiming patent rights. The vast vast majority of the 0.1% of software that does get patented—and quite possibly all of it—would still have been created if there were no patent rights to software. Companies file patents on software they happen to have built for other reasons; they don’t build software to file patents.

Software patents are “fundamentally broken” because they do harm, and they provide absolutely no benefit whatsoever. As has been pointed out so often, the monopoly rights granted by patents actually stifle innovation, and in software there is no increase in innovation to even slightly counterbalance this. The contribution of software patents is net negative to innovation.

When Google writes a letter to Congress demanding legislation to prohibit the patent office from issuing any software patents in the future, and invites other large tech companies to sign it, then they’ll have claimed the high ground on the patent issue. Until then, they’re just whining because they lost the last round in a game they’re perfectly happy to play.

Update 2011-08-16

At risk of sullying my opinion with facts, I’ve stumbled across a working paper detailing a few actual numbers for software patent filings.