On Mon, 2007-03-26 at 08:29 -0500, Mike McGrath wrote:
Thorsten Leemhuis wrote:
BTW, did anybody ask the moin moin developers for their plans? I suppose other projects hit the same bottleneck, too...
Yeah, I've talked to them a bit. Toshio is actually working with them to submit a patch.
... but I am short of time. So if someone wants to work on this first, the basics are that we want to maintain a list of subscribed pages. One list will be per page and have all the normal subscriptions which target one specific page. The other list will be global to moin and hold all the regular expression using subscriptions. These lists only have to be updated when the user updates their subscription preferences. The Moin code will then be able to quickly assemble a list of people that are subscribed to the page from the per-page-list and sequentially trying to match the regexes.
(The Moin devs tell me the current method scans every subscription request of every user to see if it matches the page that has been modified. 8000+ users of which 1000+ have subscriptions. In the new scheme we only have to sequentially scan 180+ regular expressions.)
I don't think I'll be able to work on this until post-F7. So you guys have nearly 2 months to get fed up enough with the delay to get to this first ;-)
-Toshio
infrastructure@lists.fedoraproject.org