[jcranmer]

A long time ago I was vehemently against all software patents. But as I thought about it more, I ran into the same problem you did: where is the boundary between a clear "machine" patent versus a software patent, especially if you look at modern microprocessors that rely heavily on microcode.

My resolution to the conundrum is this: software should be patentable, but we should recognize that the bar for a software patent is much higher. It is much, much easier to iterate on several ideas and concepts in software than it is in hardware: I can try five approaches in a day in software, but building five prototype machines may take several months.

As an example of something that might make the bar, I'd submit PageRank. It's easy to say that it's merely taking the eigenvalue of the Internet, so therefore it's not patentable. But the novelty and creativity is not in figuring out how to take the eigenvalue, but recognizing that the result of this operation is tremendously useful in the problem of searching the internet. The combination of the algorithm and the use it is put to is the basis of the patent--for prior art here, note that drug patents do not patent the molecule itself, or the process of manufacturing the molecule, but using the molecule to treat the condition.

[jacques_chester]

The thing is, you don't need to build a machine to patent the design. You just need to design it. So you can still try five approaches a day for a hardware invention as you do for software.

I generally agree that picking the right line is hard and that nobody will ever be happy.

[jcranmer]

There used to be a requirement that you had a prototype in order to obtain a patent. I wouldn't mind going back to that requirement.

[wiggles_md]

Restoring that requirement across the board would have a profound negative impact on the biotech (and likely I would guess, advanced materials) sectors. Think of, for example, discovering a useful small molecule, and wishing to build a moat around it while you do the med chem work for a lead compound.

[jcranmer]

You have to at least build the compound to do the assay to actually find out that it's useful.

[wiggles_md]

What you describe is the current practice. What I am trying to suggest is that it is reasonable to be able to use a Markush claim for obvious related compounds.

[jacques_chester]

Maybe, but it's not very tractable to have thousands of prototypes being delivered to the USPTO each day. The point of the patent is to provide all the information required to completely recreate the invention and explain how it works.