How Software Patents' Fuzzy Boundaries Create Unnecessary Litigation
from the subject-matter-test-please dept
James Bessen and Michael Meurer, authors of an important new book on the patent system, have a great post on the problems created specifically by software patents. They argue that the most serious problem with software patents is that they tend to cover abstract concepts rather than specific physical devices or processes. As a result, the boundaries of software patents tend to be uncertain, leading to a lot of litigation. In many areas of patent law, the "enablement" rule (which says that patent applications have to describe an invention in enough detail to "enable" someone to replicate it) helps to ensure more precise definition of patent boundaries. But the patent office only requires a general description of an "invention's" functionality to get a software patent. As a result, there tends to be a lot of uncertainty about what a software patent covers, and uncertainty inevitably spawns litigation.
Bessen and Meurer don't offer a strong recommendation on the best way to solve the problems with software patents, but they tentatively endorse a "subject matter test" -- that is, reinstating the ban on software patents -- as one part of a solution to the problem. However, they worry that a subject matter restriction won't entirely solve the problem because applicants might resort to creative drafting to evade it. I'm not sure it's so hard to draw a line to exclude software patents. Ben Klemens has suggested a standard that strikes me as pretty serviceable: mathematical algorithms are not patentable, and coupling an algorithm with "insignificant postsolution activity" does not transform an unpatentable mathematical algorithm into a patentable machine. In particular, the mere act of loading software onto an ordinary general-purpose computer cannot transform an unpatentable algorithm into a patentable machine. Although this standard might not invalidate all problematic software patents, it would invalidate most of the really harmful ones. To take one example, NTP's infamous wireless email patents almost certainly wouldn't pass muster under Klemens's test because the "invention" in question consisted of running certain email-processing algorithms on generic computer hardware. If you took away the software component, you'd be left with an unpatentable collection of generic computers and generic wireless links. I'm sure there would be some hard cases that Klemens's test wouldn't deal with precisely, but it's certainly more precise than the tests the Federal Circuit is using now.