Is non-obviousness actually applied in this sphere? In my corner of computer science, the non-obviousness bar is so low as to be basically non-existent. There are tons of patents where A and B each have significant prior art, but the completely straightforward combination "A+B" somehow gets patented.
It is applied, and it should also be in the CS space. But, as always, there are bad patents issued because they are examined by busy humans. In these cases, opponents of the patent have a number of options, most notably asking for a reexamination of the patent. During reexam, you can bring missed prior art to the USPTO's attention in an effort to invalidate the patent.
EDIT: I should have also mentioned, another option is to ask the courts to invalidate the patent (for obviousness, lack of enablement, or any other statutory requirement) if you are involved in a lawsuit. Even if you are not sued you can begin such a lawsuit by filing a "declaratory judgement" complaint against somebody who you reasonably feel might sue you.
As of 2007ish courts rely on the "Graham factors":
- the scope and content of the prior art;
- the level of ordinary skill in the art;
- the differences between the claimed invention and the prior art; and
- objective evidence of nonobviousness.
