At Least They're Asking the Right Question

After a variety of efforts at what I've termed "downstream" fixes to the patent problem, the EFF appears finally to be turning its attention to the source of the mess, issuance by the PTO. The blog post by Daniel Nazer is titled "EFF Politely Asks PTO to Stop Issuing So Many Crappy Software Patents."

Take out the word 'software' and I'd be in complete agreement. Bad software patents have gotten a lot of attention lately but rules for reforming patent examination and issuance need to be universal. You can't just single out bad software patenting practices and ignore errors if they are happening in hardware, biotech, etc. The EFF do focus on a problem that is endemic to software patents - overbroad claiming. In most other fields of patent arts it's necessary for the invention to be narrowly described and for the patent only to protect the specific claims. For example, if I patent a medicine to cure headaches I am given protection only on the specific medicine I disclose in the patent, not on the entire field of headache cures.

The post also renews EFF's earlier calls for source-code submission, with which I sympathize but I think will make more trouble than it solves. For example, what language(s) will be accepted? And how will you prove that two source code submissions are or are not equivalent? I haven't looked lately but I think proof of program equivalence is an NP-hard problem to solve. Really, though, you don't care about the code. You care about the algorithm the code implements, and we have some pretty well-understood ways to describe algorithms without reducing them to specific code forms. Yes, it may take a certain level of skill to understand non-textual algorithmic representations but we ought to expect the examiners of software patent applications to be able to read those, just as we expect other examiners to be able to read mathematical equations, or chemical reaction formulae.