On Tuesday, Oct 21, 2003, at 11:38 US/Pacific, Poor, Edmund W wrote:

The kerfuggle touched my what? I dunno :-)

I was under the impression that EVERY TIME a user requests a page, the software has to double-check each internal link for the presence or absence of the linked page.

My ancient Greek cry of jubilation only applies, if this is the case...

Well, I generally recommend reading the software in detail to see what it does ;) but here's a summary for those who wish to remain partially sane:

* Grab page info from db * Are you anonymous user? if so: * check for cached output html * compare its timestamp against cur_touched * if ok, send that out and leave * Check if client told us it's got a locally-cached copy of the page * compare its timestamp against cur_touched * if ok, say to use the cached copy and leave * Do a quick query to the link tables for all existing and non-existing links from this page * Parse the wikitext * As we find each wikilink, look it up in the lists we already got to see if it exists * On the off chance the tables are corrupt, anything that wasn't listed will be checked individually against cur * Output * If anon, save output html in the server cache.

On page creation and deletion we update the cur_touched timestamp of all pages that link to the given page, thus forcing them to re-render, and we update the link tables to reflect the new existence or non-existence of the page.

-- brion vibber (brion @ pobox.com)