(This post is somewhat of a followup to Yehuda’s post, though coming at it from a different angle.)
I agree with Yehuda that full versions of papers are, generally speaking, extremely important. As has been discussed elsewhere (see here and here, for example), our community (both the TCS community as a whole, as well as the theoretical cryptography community in particular) seems to have a problem with its emphasis on paper quantity rather than quality. (Though the quality of the research is also at issue, I am speaking here particularly of the quality of exposition.) I think also that a lack of full versions hampers scientific progress in the long run: it is a barrier to entry for others, and allows mistakes to happen more frequently than they would otherwise.
In contrast to what others have said, though, I am not sure journals are the solution. What follows is one personal story that illustrates some of the problems.
Journals: What Can Go Wrong
I recently had a paper accepted to a journal, and it will be published next month. This paper was submitted in August 2004. So it took just shy of 5 years until the paper was accepted, and it will have been over 5 years between submission and publication. It wasn’t as if the paper needed heavy revision, either — the reviewers just sat on it for a very long time. Nor did I get any useful feedback from the reviewers: I’m not sure they even checked the proof (whether they did or did not, they had no comments about it), and the only substantive thing they made me change were the references to prior work. And I didn’t even agree with their recommended changes (but didn’t feel like fighting them either).
Because of the journal’s policy of having “new results”, I added results that (in my opinion) increase the length of the paper without increasing its quality. The journal’s typesetting rules made the paper even longer and, in my opinion, more difficult to read (due to font issues).
Ideally, I would have liked to maintain a different copy of the paper on my webpage, written the way I like. I know some people do that, but in practice it is a pain to keep track of different versions (and so I didn’t do it in this case).
The Future Role of Journals?
So if journals are not the answer, what is? Let me be first clear that I think journals do have an important role: they are immensely useful as a form of peer review (in terms of both correctness and level of interest in the results) — assuming reviewers do their jobs. We are also not going to more away from a journal-based system any time soon, not least of all because journals publications are still a part of hiring/tenure/raise/etc. decisions. Yet is also seems clear that journals will eventually die out, at least effectively, and probably within my career; I plan to say more on this in a future post.
Still, we don’t need journals, given the fact that we can post our papers on our webpage and put them in public archives.* It seems, then, that the most useful role of journals is to serve the role of being the “carrot” that motivates people to write full versions in the first place.
*Or write a book. See Goldreich’s intriguing essay: On Our Duties as Scientists (personal version), Section 4.1. His essay is worth reading for the other points it makes as well.