1- immediately availability of the manuscript
2- a transparent review process
3- a high standard of quality.
4- low cost of publication
The immediately availability of the manuscript is important because publishing in a peer-reviewed journals can often take more than 1 year - so this is problematic when making claims about who did what and when. A transparent review process is important to create an incentive to carefully formulate constructive reviews and weed out spurious reviews dealt under the veil of anonymity. In addition, we can establish the reviewer quality and reliability - important when asking for future review requests. Another other other aspect of open review is to enable the acknowledgement of reviewer contributions to improving the quality of the manuscript. In this respect, Semantic Web is stellar - it acknowledges the editors and the reviewers on the front page of the published manuscript.
One concern was whether reviewers would be sufficiently forthcoming about the failings of a paper such that it lead to the strong rejection of the manuscript. From all indications (see the paper that Pascal and Krzysztof wrote http://t.co/twDASAN9), the journal not only rejects a significant number of manuscripts that don't meet expectations for publication, but also does so in a constructive manner such that it invites authors to submit a revised manuscript with indicating how they have addressed the reviewer's comments. Indeed, this aspect of open review also means that it authors attempt to shop the failed manuscript to other journals, a simple web search should uncover the reviews associated with it. This saves precious editor and reviewer time and really pushes the authors to make substantial changes.
While this is all good, the major point of content is the business model. From any which way I look at it, I get the feeling that the tables are turned upside down. First you have the authors creating intellectual content, which they submit to a journal, and is reviewed by people who don't get paid to do so. The journal then turns around and tells the authors that if they want it freely available to the public, then the authors should cough up the money. This is ridiculous. Even a simple advertising based model could easily make a return on investment for articles of increasing impact. The other aspect is that the publisher needs to cover the costs of a print publication, but really, who reads print? I haven't for over 10 years now! In the case of the Journal, they exact a cost in typesetting, but we're used to doing this for our workshop and conference submissions - so we really don't need people to do this anymore. It would be nice to see a break down of costs for digital publishing systems today (perhaps you can point us to one).
So what is the business model of the modern scientific publisher? Well, i think it lies in aggregating and creating new content, whether editorials or comprehensive reviews, which give readers a viewpoint or summary of where the field is heading. Perhaps low-cost subscriptions (e.g. $100/yr) or pay per view ($5), one could recoup costs if the work is sufficiently meritorious. It's definitely time to think about the next evolution in scientific publishing.