Likewise, archive.org does as well. I expect these papers will eventually end up in both these places.
Academic journals are full of incredibly specific knowledge. While some of it is accessible to people with a moderate amount of intelligence/diligence, any journal whose name isn't "Science" is designed for and by an incredibly small group of individuals with an incredibly specific knowledge base. Where is the donor base for that?
Answer: Universities are the only plausible donors. And they seem to think maintaining their place of privilege in being able to access the articles is worth some extra bucks.
But I think there is not a lot of interest in changing the system. The people who could change things are not directly affected by the disadvantages of the current system.
Don't be so sure. Some experiments can produce large data sets, which ideally should be made available alongside the published paper.
But yes sharing with large academic datasets is a real and only partially solved (IOW not solved at all) problem.
http://strategy.wikimedia.org/wiki/Strategic_Plan/Movement_P...
http://en.wikipedia.org/wiki/Wikipedia_talk:Lamest_edit_wars
Of course, a case could be made that university libraries should mirror all public domain digital documents of sufficient interest. Which shunts the cost onto the taxpayers.
Regardless, the cost still is non-zero, and someone has to bear it.