The fundamental idea is solid, but there should be no need for a separate project. It is really just a matter of good research practice. Defining a "good" reference is not always a productive exercise. We can end up with NPOV disputes and edit wars just as much over the validity of references as over content.
It should be a separate project, it has its own technical challenges and it is clearly a separate database, since its content is more restricted, and its format different. More over, there is a need for lineages of sources in a way which there is not among wikipedia articles. It might well be piggybacked on wikisource - but the concerns are different. The major technical challenge for wikicite would be creating cards, creating descent for editions of the same book, and for then annotating those cards. As for administrative divisions, that I am not clear on the relative merits of creating a separate people structure for it, so will defer to others who know more than I, which on this point is almost everyone.
It would be integrated in with wiktionary, wikipedia, wikiquote and wikisource - each one having citation needs - and therefore a community of users who would want to have access to a catalog, and to annotate the reliability of sources in it.
This process will not end the need to search for consensus, the problems of POV writing. What it will do is tie information in wiki projects more closely to the sources, and therefore make users and editors more capable of making decisions about those sources. After all, the scholarly community has arguments, discussions, conversations and fights over the validity of source material, we aren't going to end that.
What we can do is making it much easier for people to have well cited and comprehensively referenced articles, which tie information together in ways that allow people to make judgments. By lowering the amount of work to reference and cite, it will make new articles better cited, make it easier to provide citations for existing articles, and allow the body of published knowledge to be at least linkable, even if the actual text is unavailable.
At the risk of stating the obvious, journal articles are shorter than books. One Wikisource contributor has recently begun work on a 1917 National Geographic issue. That should be an opportunity for getting some of the bugs out of that approach, and developping standards for the way we enter journal articles. Some journal articles may be more important than others to include. but the simple fact that they are each individually shorter may be an encouragement.
But far more are published. Having a citation available for every article published in every journal is probably out of our reach. However, if we could get some kind of feed from the major journals, that would dramatically improve the quality of citations on wiki projects.
Wikicite might be able to keep track of all things published. It would also be useful to keep track of books and their printing date to know when things enter the public domain to know when to put them on Wikisource/Wikicommons.
There is already a considerable overlap between Wikisource and Wikipedia over the matter of bibliography. For now I see it as fair game for both. I believe that Wikisource should be carrying the fundamental information that helps in determining the copyright status, but that's only going to be there if people put it there. Whatever one might think of long copyright terms it is still much easier to calculate expiry based on the year of death than on the year of publication. It's going to be another 40 years before that becomes the norm for US publications.. For them we still need to consider such issues as copyright renewals. A 1923 US publication whose copyright was not properly renewed in 1951 is now in the public domain. There is no need for a new wedge project somewhere between Wikisource and Wikipedia.
Ec
Wikicitations would not deal with the copyright issue, since the citation is a card. That card might include the original source, if available to wiki, but need not, the work of determining availability should be wikisource, which wikicite uses. The best analogy is that wikicite is to produce a card catalog on line, and a system by which citations are made available live to other.
"Fair game for both" means substandard implementation for everyone. Textual apparatus is fundamentally a different process than either writing articles, or providing correct sources, instead, it ties these two together. Currently there is no mechanism for providing textual apparatus to written sources in wikiprojects - which is to some extent why people often over-rely on web sources, because it is easy to link to them. Lower the barrier to citing paper works in articles until it is no higher than linking to a web site - that is search, copy, paste - and this problem will be fixable with education and leading by example.