As much as I don't like saying it, I think Elsevier are correct in what they are doing. If an author signs a copyright agreement that prohibits uploading a publisher's final version, he or she should uphold it in my opinion. The real problem is that copyright agreements are typically very unfair on authors in the first instance, and it is this that needs to be changed. From my experience there can be a lot of confusion over what level of sharing is allowed by authors, with some believing they are free to share 'their' articles where they like, as it is 'their' work. This, however, does not make breaching such legal agreements and contracts 'right'. Hopefully, the biggest effect of the Elsevier TDNs is that authors will now understand how restrictive publishing contracts are and the rights they are giving up, and henceforth renegotiate as far as possible and maximise what they can do to disseminate their research through legal means, such as institutional repositories.
In truth, I believe the Academic.edu TDN situation is a good great news story for IRs, and ultimately, sustainable open access. Whilst sites like ResearchGate and Academia.edu may provide a platform for authors to share their work and their simplicity may be appealing, there is no guarantee that these platforms will exist in a year's time, nevermind five or ten years' time. More importantly, open access is not the raison d'etre of these research networks, but merely a convenient by-product which has garnered them good-will from users. Researchers also need to remember that by using such sites they are willingly providing their data and analytics to a for-profit third party site to resell. In reality, these sites may actually be holding back open access by reducing green repository deposits and providing a sticking-plaster solution as regards accessibility, that has just about stopped the issue from boiling over. Until now.