This is the part of the software patent I've never understood. Surely if the standard of "non-obvious to a person of ordinary skill in the art" was applied sensibly then most of the more absurd patents wouldn't be granted.

Of course there is the 'obvious in hindsight' problem but this test has been weakened beyond all recognition.

We don't even need to look to that subjective standard. We grant people exclusive rights or patents for sharing novel techniques with other people skilled in the art, yet I have never seen a professional software programmer look to patents for information or algorithms. It just doesn't happen.

So the tradeoff here is exclusive rights for zero public benefits. That is not the way it was intended.

I work on compilers, and while I don't read patents, I read papers which explain patented ideas. I think it is reasonable to assume that some of compiler-related papers from, say, IBM, wouldn't have been published and kept secret if IBM couldn't file patents before publishing papers.

So "programmers don't read patents" does not prove "zero public benefits", because I think it is more like "programmers do read materials which wouldn't have been published if they couldn't be patented".

Are some of those papers also defensive publications to prevent others from patenting the ideas?

You wouldn't necessarily read the actual patent, but it could happen that you read a conference paper, or attend the presentation, and then learn that the technique described in it is the subject of a patent application. The paper will still be a better source of information than the patent.

I don't know how often this happens, but I have seen it.

I used to work with software that used SIFT.

And I am sometimes a programmer.

Yeah, one would think that a widely infringed patent was a good signal that it was already obvious to those skilled in the art, but the patent bar seems to have this fantasy where people are reading patents and cackling gleefully stealing all the ideas from them. So they see it more as "massive business opportunity" than a signal that the system is not providing the alleged benefit to the public.

And to be honest, from those I've discussed it with, they are largely concerned with monetization and don't give a damn about any sort of public benefit.