A natural way that one might hope to bring about a genuine change to the current subscription model where libraries pay through the nose for journals is that (i) we all put our papers on the arXiv and (ii) the libraries conclude, correctly, that the benefits from their very expensive subscriptions do not justify the costs. Bundling across subjects makes this a lot more difficult of course, but it seems that some institutions in Germany do not subscribe to the Freedom Collection (see previous post for a definition), which makes it easier. And now there is an example. The Technical University of Munich mathematics department has put out an announcement that it will cancel all its Elsevier subscriptions by 2013.
Please, if you are considering submitting a paper to an Elsevier journal without putting it on the arXiv, think of the faculty members of TU Munich who will not be able to get access to your papers (or at least not conveniently), and change your mind. If you do, it will also make it easier for other departments and libraries to make similar decisions.