It's true (AFAIK; IANAL) that if you are reading someone's source code and writing your own based closely on theirs, you may well still be infringing their copyright.

But the parent has an valid point nonetheless. If you do a clean room reimplementation of a program, then you are off the hook as far as copyright goes. Yet the user may not be able to tell the difference. In short, the user cares about the functionality of the program, not its expression (the source code).

I think that's a valid argument that copyright protection is not entirely sufficient for software, and that patents could have a useful role to play.

Personally, I think the problem with software patents today is that the obviousness bar is far too low -- we're patenting problems, not solutions. See my comments elsewhere on this page.