It may be true that people come up with algorithms and mathematical constructs without patents as an incentive, but I still don't see how the same incentives (patents) are necessary for people to build novel material goods if they are not necessary for people to build novel software goods.
What I'm left wondering is, what is it specifically (for people who hold that some patents are OK, while software patents are not) that makes software different?
Edit: I just saw dstorrs comment, and I think what he seems to be implying is that patents on software are not comparable to other industries because of the relatively lower barrier to entry, and the relatively higher cost of enforcement. I think that this is a good argument for holding software patents to different standards (i.e. shorter statute lengths and more specificity in the application) than pharmaceuticals, etc. but I still don't see how patents are entirely misguided in the case of software.
Software is different because it's considered a creative work and therefore falls under the veil of copyright. Copyrights protect each implementation individually, and this is ample protection.
Patents generally cover a whole _invention_, not an implementation, and each implementation requires a license from the patent holder. So, in the case of the web browser, had Mosaic or Netscape patented this concept every competitor would be forced to obtain a license from the patent holder and open-source or free browsers would be impossible. Patents in the case of software therefore have a chilling effect and prevent significant competition for a long time.
Pharmaceuticals only get patented by publicly divulging their significant modes of operation; a software patent outlines only vague processes, like the i4i patent that's brought this whole thing up.
Software is also merely a high-level implementation of an algorithm. Algorithms are non-patentable and all math is discovered, not invented; this has been settled earlier, and is the reason that algorithms alone are not awarded patent protection. Software is no different.
In software, the implementation details are significant enough that patent protection is not only undesirable but in fact crippling to the industry as a whole. Take a look at some software patents; in fact, take a look at i4i's patent specifically. It's applicable to basically that edits XML documents, whether it be a word processor, an IDE, or anything else. These things are way too vague and they encroach on prior and future art in a hugely negative way.
See http://w2.eff.org/patent for a list of some rather absurd software patents granted by the USPTO; even if we ignored everything else and assumed that software patents were OK for things that deserved them, the people in charge of patents don't have anywhere near the kind of knowledge necessary to make intelligent decisions.
Those are my reasons for opposing software patents. It makes development a complete minefield, and only the big guys can afford to wade through that. I just have to hope that I don't do anything too successful until the patent laws are revoked or until I make enough money not to care.
What I'm left wondering is, what is it specifically (for people who hold that some patents are OK, while software patents are not) that makes software different?
Edit: I just saw dstorrs comment, and I think what he seems to be implying is that patents on software are not comparable to other industries because of the relatively lower barrier to entry, and the relatively higher cost of enforcement. I think that this is a good argument for holding software patents to different standards (i.e. shorter statute lengths and more specificity in the application) than pharmaceuticals, etc. but I still don't see how patents are entirely misguided in the case of software.