It's not a matter of "shoulds" - but rather one of cause and effect, incentives, and careful balance. The original concept of patents was borne out of traditional engineering fields, where the costs of bringing something to market is very high, but the costs of simply copying someone else's innovation is substantially lower.
This disincentivizes inventors and gives a large amount of market advantage to copycats, and when this balance reaches a point where inventors cannot reasonably expect to profit off of their inventions, innovation stops.
The problem here is that we've taken this concept and applied it willy-nilly to software without modification. Software is a field where experimentation, innovation, and even product production is fundamentally dirt cheap, perhaps hundreds, if not thousands of times cheaper than traditional engineering fields.
In this case, applying the system as-is has resulted in a gigantic power shift towards the patent-holder/inventor, where they can hold the market hostage, and the fact that you can iterate so quickly and cheaply means that, under current rules anyways, a solid day's work is probably patentable somehow. This is obviously bull, and needs to be fixed - but I'm not convinced removing patents globally (i.e., from all fields, not just software) is a good idea.
IMHO the concept of IP must exist, even for software, but we must also account for the fact that software is intrinsically insanely cheap to produce (compared to just about everything else in life), and that the amount of exclusivity, protection, or just plain definition of what is sufficiently innovative to be protected, must change.