A number of other folks have already reviewed these policy documents, though few have offered a full commentary or interpretation (perhaps because there are just so many moving pieces here.) I found the commentary from Steven Harnad (comparing to Elsevier's first sharing policy, from 2004) particularly useful. There seem to be a couple of positive changes in the policy, but mostly it's not a huge change from things in the past, and certainly not an improvement. I can't muster the attention needed to try to read and fully parse this level of complexity, so instead I thought I'd offer just the following two observations.
An Object Lesson For AuthorsThis is at least the third version of Elsevier's article-sharing policy in the last 10ish years. I also know of other publishers who have similar policies that have also changed over time. These policies let Elsevier, and other publishers with similar approaches, obscure the fact that they require authors to give them the copyright in these works; i.e., that authors no longer own these works.
There are a lot of reasons why publishers need to have some of the rights in the work. There are also a lot of reasons (administrative-overhead-wise) why publishers don't really want to negotiate different rights agreements for each individual article. But there are few reasons other than profit why publishers need -exclusive- ownership of rights. It would be pretty easy to administrate a publication agreement with standardized non-exclusive (that is, shared) rights. Lots of publishers do so already! And publication agreements, copyright agreements, transfer statements, whatever you may call them - those are -contracts-. They can't be altered after the fact except by agreement of the parties, unlike unilaterally-granted "sharing policies", which can be altered (or repealed) whenever, by the unilat that granted them.
Many authors have suggested to me in consultations that they can't possibly negotiate with a publisher, they just don't have that power. But however much any given academic needs a publication credit, the publisher needs your content! There are lots of factors that dictate how much any individual can or should negotiate about a publication agreement - but not having anything the publisher wants is not one of them.
If there's anything you want to be sure you can do with your content in the future, for goodness' sakes get it in your contract.
Why these changes, now?My guess is that the latest policy revision is an attempt to address criticisms of the immediately-previous version of Elsevier's sharing policy, which allowed authors to archive their final edited manuscript in an institutional repository but only if their institution did not have a policy requiring authors to do that. That was... okay, really, just weird.*
I think this policy is also intended to address the proliferation of article-sharing by authors via for-profit "academic social networking" sites like Academia.edu and ResearchGate. One piece of supporting evidence is that Elsevier is now requiring that certain shared copies bear a Creative Commons license that contains a Non-Commercial clause, limiting their usefulness for for-profit enterprises. The fact that the license also contains -other- limitations (BY-NC-ND is what they're requiring) that few non-profit archives would really want to support is perhaps just icing on the cake.
I actually have a lot of sympathy for publishers' frustration with the sharing that goes on on those sites; those sites are competitors for the for-profit publishers, just as much as we non-profit archives are. They are also more appealing and easier to use than many of our open non-profit repositories - for a variety of reasons, including that they do fewer annoying things like making authors really check whether they have a right to share. So we have less content than the social networking sites, even though we've been at it longer, and even though we are not intending to monetize our users' data or research networks or anything like that. Personal frustrations aside, though, authors who want to share articles are going to do so on whatever platforms are most appealing to them.
So I think the new "Hosting Policy" is also part of the attempt to address sharing on for-profit sites. Trying to set a policy that affects the behavior of entities with whom you are not in any kind of contractual relationship is... kinda weird. I am pretty darn confused how this hosting policy is intended to interact with the policy on author sharing. If an author thinks they can share, they are legally allowed to do so, and hosts are legally allowed to accept shared copies. That's even true of the for-profits; we just all have to take them down if it turns out the author was wrong. With this new hosting policy, Elsevier seems to be trying to insert another note of uncertainty, at the very least.
In addition to criticisms of older versions of their sharing policy, and a reaction to newer players in the article-hosting arena, I think this revised policy - and the spin that it enables more sharing - reflects growing traction for open access principles in the academic world, and more public focus on related issues. I hope.
*This may not be an exactly accurate characterization of that version of the policy. I have never actually been able to explain it in a way that makes sense to any of the authors who have consulted with me, so...