Sunday, March 02, 2014

Obviousness in Software

Bob Purvey (a friend who is both a software engineer and a registered Patent Agent) has written a fairly short paper to explain why the patent office doesn't do a good job screening out "obvious" software patents. He observes that the approach they use for patents on pharmaceuticals seems to do a much better job, and shows how it would apply in software.
One of the reasons this is relevant is that whenever the software industry tries to get the federal government to reform the patent laws (since they're so obviously broken as applied to software), congress is counter-lobbied by the drug companies who argue that patents are crucial to their business model, and without the present system, new drug discovery would whither away. Congress, of course, doesn't understand either industry, so they collect contributions from both sides and do nothing. If the software industry had a more focussed proposal, it would be more likely to get the kind of change that would be useful to it.
Purvey's argument is that when deciding whether an application for a patent on a new drug or medical technique is novel, the examiner is expected to consider whether a Person with Ordinary Skill in the Art (POSITA) would think of that solution, given a suitable description of the problem. If the approach described is one of a handful that the POSITA would think to try, ("obvious to try") then it's obvious enough to be ineligible for a patent. If it's one of a thousand approaches (in the drug business, this is now common in drug discovery) that you'd have to try, with an unknown likelihood of success then it doesn't count as obvious.
The patent office often grants patents on well-known software techniques applied in new contexts. Purvey argues that the "obvious to try" standard would invalidate those patents because software people are trained in abstraction, and it's obvious to all skilled practitioners that previously known techniques are likely to apply in the new context once you describe the problem correctly. The fact that the patent applicant described the problem with non-standard terminology doesn't invalidate the standard tools, and that should be an acceptable argument when suing to invalidate a patent.
As Purvey showed, that's an accepted standard in drug patent trials. A prominent (pharma) case he talked about in the paper hinged on showing that the patent application used non-standard terminology to describe something that was an obvious combination of published techniques. If lawyers attempting to invalidate a patent brought this standard into play more frequently, more patents would be invalidated quickly. And this is an approach that shouldn't have to wait for new action from legislators.