Imaging what you could do with a social-network like publication database (research-gate sort of is, but it doesn't think big enough and the UX is confusing, imho). Every publication became a "post" so to speak. You could easily include discussions (both on the paper, as well as general discussions), ratings, user curated lists (instead of expensive journals), host links to any relevant non-paper data. Funding could be provided by Universities, or by means of a registration fee, or simply with an ad supported model. Targeted ads will probably bring in a relatively high CPM due to the target audience.
<rant> Incidentally, while we're at it, we could also get rid of strict formatting guidelines and super-formal writing, some of the best papers I've read are informal. And it's a very modern thing to do anyway, old papers by Kelvin and the like read much better than what we have to write these days. </rant>
For any success of such a project, the official endorsement from arXiv is crucial. Academia is a very conservative place; anything branded as a social network might hit a snag, especially if "pricing" or closed source recommendation systems pop up.
[1]: http://www.researchgate.net/
[2]: https://www.authorea.com/ -- It costs quite a lot if you have several papers and you wish to release them only after the preprint is ready.
I don't think official endorsement by arXiv is necessary, what is important is to gain a critical mass. I think authorea has the right idea to give users a incentive to use the program. If possible, I would offer more free access to gain that initial traction, but of course this requires an investor with deep enough pockets to take the hit. Networking would also be key, to bring it back to the original article: if you could work with those involved here and offer your services to host all these open-access papers...
>Academia is a very conservative place; anything branded as a social network might hit a snag, especially if "pricing" or closed source recommendation systems pop up.
I think we need something similar in openness and positioning as Github. AFAIK Github did not market itself as a social platform at first, but it just made it so easy for its users that it ended up being key to its success.
I think that is the overall approach one should take, don't force the users into a pattern, that will not work and definitely not in rigid-academia. You have to guide them by a good UX, make them think they want to use your features because that is what's good for them (they're like cats in that regard).
Authorea is very different; it has a very tough pricing model where you have to either choose to publish even partial work while you are writing (unacceptable to most scientists I know) or pay per article AND per month.
Basically, my issue with it is the same as with the current open-access model: I am trying my hardest to produce good science and release it for free; yet somebody tries to profit not from my own results, but from my desire to have those results published for free. ArXiv helps me while refusing to profit from this desire, and thus I love it.
You have to keep in mind that a lot of scientists (I think this is true for compsci/math/physics) choose a 50% pay cut so that they can do science, and not go into the (programming) industry. Speaking as a PhD student, we have modest means and do not like to be gouged.
Drafting, extending and version control would be particularly important, I think. It's been a long problem with dead-tree journals, following adjustments in results is hard because they inherently rely on back-links, retractions (in extreme cases), letters, etc. Not a coherent way of centralising information.
I don't know how you'd do useful public peer review. By definition, "peer" assumes some professional domain competence.
The current system is probably too restrictive, but it's hard to imagine a more open system that's at least as useful.
Maybe someone should do a PhD on it. :)
Heck, you can even create groups like "former reviewers of XXX journal" and as soon as you convince other people, gain reputation very quickly.
I would also make reviews public and name-and-shame any caught review fixing (if proper evidence is available), which would be the bigger problem imho. But again, I think this is something we should also be able to detect using proper algorithms. If a user Alice always rates the papers by Bob highly, even when others disagree, perhaps her opinion of Bob's work is not to trustworthy.
This is how some journals come to be more prestigious than others anyway. You don't restrict the creation of journals to verified people; instead everyone gets to run a journal and they compete for reputation.
OTOH, I do not like excluding users, but I think public read-access and write-access trough accredited universities and research groups would be a fair balance. (Note, I also think if there is a charge associated with this process we should charge relative to the users country. Subsidizing access for developing countries by charging more in developed countries. To reduce exclusions.)
Note that arXiv has features that many previous repositories did not have, such as easy access to the original TeX files.
Plus, a comment system for a repository of preprints sounds very useful to me -- you can have early birds (such as PhDs or very active scientists) comment on your paper even before you submit to a conference/journal, thus getting good advice for free.
Anyway, "it does not have it because it has not happened yet" is exactly my point, because I don't see why it cannot evolve into a publishing platform.
It's only really used by the quantum information community at this point, but you could try promoting it in yours.