Physics is lucky in having a single “journal” in which nearly all new research is published – the eprint arXiv founded by Paul Ginsparg.
A problem with this fabulous system is coping with the firehose of information it produces. On a typical day more than one hundred new papers appear. Just reading the titles and abstracts would consume a fair fraction of your day.
How on earth are people supposed to cope with all this information?
1) With a daily, automated, relatively dumb search on keywords, which I suspect is already implemented.
2) With a reviewing system, which as part of its functionality, allows you to choose which reviewers’ marks you pay attention to. Ideally, a multi-criteria reviewing system with independent marks for novelty, interest, writing style, etc.
3) With an agent system that merges those two functions and applies some automated bayesian statistics– the reverse of the bayesian spam-killers, perhaps– that taps the universal journal and filters it into your personal journal.
Scoring/reviewing parts of arxiv.org has been tried without much success before (e.g. Daniel Gottesman’s initiative).
One step is to use a standardized digest format such as RSS or NNTP. I am personally using arxiv->RSS->NNTP via http://hein.roehrig.name/arxiv and nntp//rss http://www.methodize.org/nntprss/
HTH,
Hein
Making the review process open, e.g. by having everyone blog their reviews, might enable collaborative filtering and make interesting things much easier to find – just as is happening now in weblogs in general.
In addition, weblogs as review-collections-by-a-single-author have the advantage of potentially bringing visibility and recognition to reviewers, which could help improve the quality of reviews.
Right now, as Doron Zeilberger wrote (http://www.math.rutgers.edu/~zeilberg/Opinion25.html) , “referees are anonymous, and can afford to do a careless job.”