On Wed, Feb 16, 2011 at 6:55 PM, Carcharoth <carcharothwp(a)googlemail.com> wrote: Make sure you read the history too, where I removed all the bogus technical claims three different times over the span of a few months before giving up when Wikid77 re-added them a fourth time. All the claims that links per se are a significant technical problem are complete garbage. The links tables are large, but not any kind of bottleneck. The thing that makes parsing slow is massive amounts of wikitext, whether it's links or anything else. Deeply nested templates might be a particular performance issue, for instance, but links table storage is definitely not. But I've given up on trying to keep Wikid77's persistently uninformed arguments off that page. On Wed, Feb 16, 2011 at 10:03 PM, Tim Starling <tstarling(a)wikimedia.org> wrote: AJAX would lead to a poorer user experience, since they'd take some time to open. Wouldn't it make much more sense to have some kind of HTML-level cache for templates? Normally you can't do that, because wikitext constructs might start and end in different templates. But perhaps templates could opt in to an HTML cache through a magic word to get improved performance, at the expense of behaving somewhat differently. This should work for the giant static navigation templates as long as they don't vary per-article in important ways. Although of course, you'd need someone who understands the parser to write the feature, while AJAX could be done by any Wikipedia editor who knows some JavaScript. (This is clearly more of a wikitech-l discussion, though.)