The open access journal collection BioMed Central is launching a very different biology journal called Biology Direct. Not only will the publication be open access, but the reviews of the paper will also appear, unedited (even if they're harsh), alongside the data.
The traditional way of publishing is that an editor referees the interaction between the submitting scientist and the reviewers, and makes the final call about whether something will be published. The costs (editor's salary and endless secretarial support) are generally recovered in part by subscriptions to the magazine in question, but there have been increasing complaints that this shuts out the public from work which is after all tax funded. Open access journals let you download all the publications for free, but still have the editor infrastructure, which is paid for in part by a fee charged to the researcher. I guess Biology Direct (which again charges the researcher) would fall one step closer to self-editing communities like Wikipedia, in that there is a in house stable of reviewers (the Editorial Board) who have to be approached by the scientists in order to get a review. If you get a review and you pay your dime, apparently, your work goes up on the web page.
This seems to put a lot on the Editorial Board, and also I wonder if reviewers won't pull punches since they know the full text of their critique is going up. I took a look at the most recent accepted paper, Glazko et al. (link is to abstract; the article, reviews, and responses to reviews are in one large pdf) and the reviewers did seem to have gone over the paper pretty thoroughly. It looks like revisions are put into the main body of the paper without special designation (this is the practice in more conventional publications).
A good statement of Biology Direct's working philosophy and rationale is here (keep scrolling down), and there's a very good and link-filled writeup of the rationale for the new proposal by Jamais Cascio over at worldchanging. Very interesting, and a bit of a risk.