On Wed, 26 Mar 2003, tarquin wrote:
do we generate the HTML from Wiki code every single time a page is viewed?
Presently yes, unless your browser caches it on your end.
To do caching on the server, we have to finish working out issues of changes made to the html output by variations in user preferences (some things, like the TeX html/image/mathml? options, will perforce alter the output; others can be relegated more to the stylesheets than they are now) and whether we can also cache the whole thing including the template.
The fastest thing would be to have a complete, ready-to-send page as a file which we can spit right out upon determining that the requested page is cached and the user isn't using funny options. Putting username/ip addr in the corner, having a page counter, and having different sidebar links available for anon and logged-in users complicates all this.
Another thing; on phase 2, we had problems with the caching code wherein some things that were determined from parsing the page (interlanguage links and link keywords for a meta tag) got dropped when viewing the page cached. These either need to be pre-cached too, or need to be separately fetchable. (For instance, magnus's experimental code with a separate table. This needs to get finished and put in place!)
See http://meta.wikipedia.org/wiki/Cache_strategy
-- brion vibber (brion @ pobox.com)