The big problem that keeps coming up with software patents is the fact that most of them are bogus, because of being trivial or unoriginal. Barring cases of malicious intent by patent examiners, which are a completely separate issue from the viability of the patent system itself, the reason for that is that the patent examiners don't understand what constitutes non-obvious work in software. In fact, I would guess that now there are very few fields in which it's feasible for someone who's not already familiar with the field to make that judgment in any reasonable amount of time. Since the criteria for patent validity are based on how the work would be viewed by average people in the field, maybe we should put patent-granting into the hands of people in each field, who are actually qualified to evaluate such work. Just have the central patent office keep records and oversee devolved organizations for each discipline, who actually control which patents are granted and what the terms of patents in that field are. That way, not only can bogus patents be avoided, but fields with radically different environments won't be hampered by patent terms designed for other areas (e.g. software patents could have a much shorter duration than others).