(Brion Vibber vibber@sal-sun062.usc.edu): On Wed, 26 Mar 2003, Nick Reinking wrote:
Speaking of this, I don't see why they can't be reenabled. However, instead of doing it in realtime (would which kill your performance), why not do it at night? Parse through the Apache log, compile a list of how much each article counter needs to be incremented, connect to MySQL... and do it. :)
Because I don't care about the counters, and I'm still way behind schedule on things I think are important, like better caching and setting up additional languages on the new system, so I haven't gotten to it; and no one else has done it either.
Here's what a line of the wiki's logfile looks like: 20030326195656 0000.175 /wiki/Styx timestamp response time path
Parsing is left as an exercise to the reader... :)
While we're at it, I added the wiki logfile just to get some initial stats, but do we really need it anymore? I'm sure Apache is much more efficient at it, and it would eliminate a disk access per page fetch. The test suite can be used to get timing info now.