On Wednesday 15 August 2007, Patrick R. Michaud wrote: > > > How about this: We can also design a special sort of super-short-time > > cache, which could store a HTML output (for non-identified visitors) for > > 5 minutes, and it could send this cache to the visitors without any other > > checks of permissions/last-modified pages/pagelists. > > IMO, using time limits on caches really isn't useful here. First, it > greatly reduces the odds of a cache hit (i.e., we only receive a benefit > when a page is hit more than once every X minutes).
True, but if we do have less than one hit every X minutes, even old/slow servers could handle the load. > Second, it's not > needed, since we already have superior methods to determine when the > cache is invalidated (i.e., we can invalidate things based on a new > posting). I meant this way we can even eliminate a stat() on the second file (.lastmod) during the slashdotting, and only do it once every 5 minutes. > > > This may even be done before calling pmwiki.php: just checking for > > our cookie, and if there isn't a cookie, and the cache is newer > > than 5 minutes, send the cache. > > Which cookie? I was thinking of the cookies PHPSESSID, $CookiePrefix.'author' and $CookiePrefix.'imstime'. If they are not sent by the browser, the visitor is very probably a non-identified reader and any (:if auth edit:) and other personnalized content are not applicable to him, so, the output of PmWiki for him is cacheable and reusable. If there is a valid cache, we can even not at all call pmwiki.php. Petko _______________________________________________ pmwiki-users mailing list pmwiki-users@pmichaud.com http://www.pmichaud.com/mailman/listinfo/pmwiki-users