Another point worth noting is that many (should be all) papers are readily available (legally) as preprint PDFs (which is usually almost the same as the final version), if the creator bothers to place a preprint online, and most academics would be happy to email you a copy of their work should you ask anyway. So it's not like publishers are forcing someone in a walled garden, it's just that academics still prefer to be in there. So, these attacks on Elsevier are shooting the messenger, but it seems it's the only way to get to the sender here.
I've seen the occasional clueless student tell me they didn't think it was legal for them to upload their own papers (they might have signed some contracts with illegal clauses to that effect for all I know). Counting on the authors to unbreak academic publishing works if you already know what you want and can deal with a bit of hassle searching and e-mailing, but it doesn't work at all for skimming or looking at complete proceedings.
I suspect the publishers are fighting a bit harder against competing archives. For example, because they are sometimes involved in organising conferences, they can entrench themselves by offering an archive to the participants (giving a small sponsorship in exchange for the understanding the organisers won't offer the same archive), locking out the rest.
