Ok, let's be fundamental about this. Isn't it strange that we should consider "software" as different from other intellectual property? If X hours of work have been invested into the invention of a clever software routine, then, it would be strange if a patent could not be granted for that work while a patent would be granted for some physical apparatus that also took X hours to develop. (Don't think about the stupid "one-click-buy" software patents, but more along the lines of an ingenious differential-equation solver).
So, I don't think a law that says "patents are granted, but not for software" would be a good one. If we would abolish patents, we should do it in all fields.
First, patents are not granted on the basis of the effort expended to invent a patentable invention. The sweat of the brow theory is just as much bunk for patents as it is for copyrights.
Second, the purpose of patents is to encourage the invention, disclosure, and bringing to market of inventions which otherwise would not be invented, disclosed, and brought to market, and where the restrictions on the public are as minimal as possible in both scope and duration. Patents have an inherent negative effect on invention, disclosure, and bringing to market, and so it is important that the incentive is large enough to spur on more of this behavior than it inescapably deters. Further, patents inherently limit the freedom of the public to practice the invention, and tend to have negative effects on the market due to the monopolistic prices the patent holder can charge, so it it is important that the positive benefits of the patent for the public also outweigh the inescapable negative effects it has on the public.
What's interesting about the software and business method fields is that there are many natural incentives for invention, and bringing to market. And while formal disclosure could still be useful, the system is gamed to make disclosures unhelpful and at any rate obervation of the patents in practice in these fields usually reveals anything that disclosure would. This means that the incentive of a patent amounts to little in these fields, but the negative effects of the patent are not mitigated at all. Thus patents here act to harm inventive activity more than they do to spur it on. Combined with the negative effects on the public, software and business method patents wind up producing more harm than good.
Someday, perhaps, the natural incentives in these two fields will diminish and there will be more of a role for an artificial incentive from patents. By all means we should watch for that so that we can revisit the issue when th time is ripe. But for now, software and business method patents harm more than they help. That's why we need to be rid of them.