Since what readers need is: 1) assurance of quality and readability 2) assurance of permenance
... all you need is a system of review and an archiving service. You do need a journal as such. Authors would put their paper on the web and give notification to the relevant subject review board that they wish their paper to be reviewed. It is then reviewed (either by traditional blind review or open) and given star ratings (zero to 5) for content and for presentation and entered on database. Readers could browse this at their own choice of level (only 5 star on both content and presentation, above 2 for presentation and above star for content) as well as on subject codes and keywords. Additionally they could subscribe to a mailing list to be notified of new papers at their chosen levels of accepatability. There would be no revision of papers as such, but authors could re-submit papers (subject to some limits) to obtain higher ratings in presenation and content by addressing the previous (and others') comments. Every five years the archive of papers (or the top x% of it) would be issued on a CD_ROM as a subscription to libraries (who would pay for this and thus cover the work need to copy papers onto the archive). A web robot woudl periodically check that the papers were at the url and delete them from the review database if this is relevant. In this way the review service could be seen as an add-on to existing open paper archives such as xxx.lanl.gov or cogprints. Since academics already do the reviewing and much of the mark-up (or a secretary in the department does it) there would be almost no running cost for this (apart from the archve service which could cover itself with a very low subsccription from libraries). There seems to be *no* need for a publisher any more.
