Via Peer-to-Peer, Ariberto Fassati in this week's Nature correspondence (sorry, toll access only):

Reviewers [of scientific publications] often make significant contributions in shaping discoveries. They suggest new experiments, propose novel interpretations and reject some papers outright. [...] It is well worth keeping a record of such work, for no history of science will be complete and accurate without it.

I therefore propose that journals' records should be made publicly available after an adequate lapse of time, including the names of reviewers and the confidential comments exchanged between editors and reviewers. The Nobel Foundation makes all its records available after 50 years, as do many governmental and other institutions. This delay may be reduced for scientific journals to, perhaps, 15 or 20 years.

Now that's a damn good idea: it's long past time that reviewing got its due as an essential part of a scientist's job, and opening the records should help to generate such recognition (to say nothing of the invaluable contribution to historiography of science).

My only quibble: why 15 years? If six months is long enough for an embargo on a closed-access paper, why is it not also long enough to keep the reviews secret? I presume the idea is to prevent retaliation for harsh reviews, but if all the information is public it would take a truly dedicated holder of a truly heinous grudge to follow up (in such a way as not to get caught doing it!) after six or twelve months. More to the point, we can dramatically reduce the risk of such retaliation by changing the community attitude towards reviewing. If peer review becomes a fully acknowledged part of the job, excellence in which is respected and rewarded -- and if everyone knows their reviews will be made public! -- then low quality (gratuitously mean, ill-informed, lazy, self-serving, etc) reviews should be a thing of the past.