My point (working in an academic digital library and just seeing the amount of thesis, dissertation, articles passing by) is that if for people is a difficult, overcomplicated burden to upload a PDF in an institutional repository (5 minutes of their time, even less), how can we wikilibrarians think that they will come to us and upload and "curate" their text?
I suspect the current population of scientists will need us to step in and perform all the roles of a journal publisher, thus allowing them to continue their workflow completely without any change to their own scientific work. If we're going to capture science publishing, we have to be MORE accommodating than the for-profit journals. So if a scientist's existing "upload" process is to just send it as an email attachment, or even god forbid to print it up and mail it to somebody, we need to be able to accommodate that with all the ease-of-use that their existing provider gives them.
Essentially, we need to steal somebody who runs one of the existing scientific journal companies (or one of their free alternatives) and get them to teach us and our computers how to do everything the professional journal companies are doing-- so that the scientists can then be free to just "swap" out their commercial publisher and switch straight to us without ANY changes.
Alec