Presently, the reputation of a journal (and therefore their peer-review process) is decided by free-market principles - those who publish the highest quality papers receive more subscription payments from users, gaining a higher impact factor and therefore higher quality submissions.
The Open Access initiative has suggested that the cost of access and review should be switched from users (libraries who are facing overloaded subscription budgets) to the authors or institutions who benefit most from publication.
In either case, in the words of Larry Hardesty: "We are now paying an exorbitant price to publishers like Elsevier to evaluate the work of our faculties".
Essentially, all that's needed is a peer-review process that is accountable, reputable and fairly priced.
At the moment journals are responsible for both peer review and topic filtering (selecting the papers that are of the greatest importance in a particular field). I think that charges can still be made for filtering and recommendation (as practised by facultyof1000 and numerous weblogs), but the mechanism of peer review has to be separated from this process.
This was the flaw in my argument last week - if papers are free to access and the index of peer-reviewed literature (MEDLINE) is also free to access, the user can't be asked to pay to find out which papers have been accepted by peer review as I suggested. We also don't want the user to have to pay for access to the papers. This just leaves charging for the filtering process (although if enough academics compile non-profit weblogs, this may become free too).
So, the authors pay for peer review and (the minimal cost of) distribution. The users pay for filtering and recommendation. Access to papers is free. The index of peer-reviewed papers is free.
The only problem: Who conducts the peer review, and how do we measure their reputation if they're not charging users for access to the results?