>>I think it's better to leave it at all software, but this still ends up with a tricky definition. What is "generally used computing hardware"?
It would be a tricky definition to get right, and there probably would always remain a gray area. But it shifts the software patent problem far away from where software patents are doing most harm.
If this limit on the effect of software patents ever gets accepted, and (at least) a single "generally used computing hardware"-platform is established, a programmer can just sit down and write software on this platform and distribute it for users of this platform.
It pushes the bickering to the edges of "generally used" and specialised hardware, reducing the problem to spats between dedicated hardware manufacturers and leaving the average programmer free to implement improved mailclients, one-click-shop webpages, and other everyday software to their hearts content.