Medaille's suggestions are constructive:  offering institutional experiments
   or reforms that might be improvements.

   Our current system is to screen out error and poor quality by peer review.
   But since journal editors and referees are not well-rewarded for doing a
   good job, the system often weeds out high quality and lets low quality pass
   through, and sometimes even takes high quality submissions and turns them
   into lower quality publications (after two years of revise-and-resubmit).
   And  as  others  in  this  thread have pointed out, it often weeds out
   creativity, and novelty, in substance and method.

   At the Summer Institute for the Preservation of the History of Economics a
   few weeks ago, one of the papers discussed the difficulty Gary Becker had in
   getting some of his innovative and controversial work published during the
   early stages of his career.  And James Buchanan in his talk, mentioned how a
   paper of his on game theory was rejected by referees for lack of formal
   expression, even though the editor (Robert Mundell) later granted that the
   paper had made an important and sound point.

   In the discussion after Buchanan's talk, I suggested that another  paradigm
   for publication would be to get a lot of stuff out there fast, and then
   correct the errors on the fly.  Wikipedia might be one attempt at this.  (Or
   Epinions, in the realm of product evaluations.)

   When Wikipedia first came out, many of us were highly sceptical of how
   reliable the information in it would be.  But what has surprised me is how
   good the average entry is; and what is indisputable is how fast new entries
   are added; and how vast the coverage of topics.  (Some of these issues are
   discussed in part of a neat book by Chris Anderson called The Long Tail.)

   Art Diamond