I'm ideologically primed to favor free access to information (especially if any tax money was involved which is true for pretty much all research) but I think a middle ground of moving scientific publishing to a nextflix/itunes/kindle model would solve a lot of issues. Access a paper -> (micro) payment to the journal (who could in return set up another (micro) payment to the actual author). The biggest issue is the pricing model (35$/paper when a researcher typically needs to read 100+ papers per papers he/she writes) and the lack of a single point of access.

Like iTunes for science. Go do it someone. I won't envy you, I'll expect you to fail but if you pull it off I'll be forever grateful.