And a proxy is not necessarily the Journal publisher's preferred method for allowing access.

At the place where I currently work, there are at least three different types of services used for different websites:

• Proxy based authentication: this is only used for certain "behind the times" websites like MathSciNet :-p
• Athens/Institution log-in: this is cookie based (I think) and is similar to OpenID. You only authenticate with your institution and your credentials are passed to the Journal/service. The website sends a request to your institution, you get redirected to a log-in page there, and get sent back after you logged in. It is somewhere more common in Europe than in the US I think. I like it better than proxy because directing traffic through the university can be slow, especially if I am overseas.
• VPN: some services uses domain/IP based recognition and can only be accessed through Virtual Private Networking.

There's a very amusing comment in the Nature page, though, which I can read for free:

In the middle of 2001 Ginsparg left the Los Alamos National Laboratory, under less than amicable circumstances, for Cornell University which allowed him to continue and extend arXiv using it as a model for research into digital libraries. As Ginsparg described it, the last straw was a recent salary review which described him as, "a strictly average performer by overall lab standards; with no particular computer skills contributing to lab programs; easily replaced, and moreover overpaid, according to an external market survey". The then Chair of the Department of Physics at Cornell, Peter Lepage's sardonic comment was, "Evidently their form didn't have a box for: 'completely transformed the nature and reach of scientific information in physics and other fields'."