I thought this was obvious to everyone? ACM, IEEE, and the rest of the scientific publishing establishment are antiquated dinosaurs. They will eventually adapt to the existence of the Internet. Thankfully, in CS, they are mostly just a nuisance -- access to the scientific literature in other fields is significantly less open.

To be fair to IEEE, the ACM's official policy is at least as bad.

Actually, the ACM are a little bit more reasonable: authors typically have to sign over copyright, but they retain the right to post "author-prepared" versions of papers on their personal web sites, albeit with an ACM copyright blurb attached (most people ignore that requirement, though).