How do software patents encourage innovation again, in actual practice?
As far as I know most arguments rely on a proxy measurement. ie since it is difficult to measure "innovation" you have to pick something else, and if you pick the right something you can "prove" your case. In the most ridiculous cases they use the patents themselves as the proxy, meaning they're just begging the question. In other examples they basically make some argument about innovation being signalled by economic growth, and then they ignore the huge confounding factors and declare that a growing economy (or growing IT sector, or whatever) is proof that software patents are a good idea. Obviously that argument is a bit fragile right now, since the economy including IT sectors is in free fall.