Last modified: 2008-12-16 01:27:21 UTC
Pages really shouldn't be able to sit un-re-rendered for years. On en.wiki, for example, there are pages with a page_touched from 2005. This means that certain tables are not populated fully (for example, the redirect table) which causes people to miss certain pages for years. E.g., http://en.wikipedia.org/w/index.php?title=Dolores_Lake&diff=210457114&oldid=68679810 I'm proposing that we force re-rendering of all pages at some point. Perhaps 30 days or maybe more, like 90 days? One proposed idea: <Vyznev> MZMcBride: sure, if we had an index on page_touched we could have, say, every edit also pull the oldest untouched page in the DB and reparse it, or something
1) There's nothing wrong with leaving pages around for a long time, that's A-OK. 2) action=purge (if you want to re-*render*) or null edit (if you want to update link tables) 3) batch scripts like refreshLinks, which aren't about re-caching.
(To clarify -- an old page_touched does *NOT* mean that link tables have not been repopulated for the given page, that's an incorrect assumption. Batch maintenance scripts such as refreshLinks and others do these repopulations and *do not* update page_touched since there is no need to purge external caches of the pages.)
