On Tue, Jul 10, 2007 at 11:30:23AM -0500, Mike McGrath wrote: > Ray Van Dolson wrote: > > On Tue, Jul 10, 2007 at 10:58:39AM -0500, Mike McGrath wrote: > > > >> I've been meaning to do this for a long time: > >> > >> http://fedoraproject.org/wiki/Infrastructure/CommonProblems > >> > >> Please add to it and make it not stink :) > >> > > > > Mike, regarding the Wiki Pages Slow Save issue... are their any bugs > > filed about this upstream that you are aware of? > > > > I will do a search, but thought you might be aware of some off the top > > of your head. > > > > I'd be interested to look at the code that does the subscription > > handling. It sure seems like this could be done much much faster, but > > maybe things are coded in such a way that it's not intuitive. > > > > Many have looked into it but AFAIK nothing has actually been done. I'm not > sure if a ticket even exists for it upstream. If you've got time please > please please make it work :) > Anyone out there run the profiler on a call to notifySubscribers() or getSubscribers()? It doesn't seem the Moin user storage system lends itself to being very fast. Each user seems to be a separate text file which must be parsed. Probably would be much faster with a sqlite backend or something, but this would obviously take a bit of work. Perhaps we could hack something in though to make the current process a little faster -- at least specifically for finding subscribers and doing notifications. I also wonder if the actual call to _sendNotification() takes a long time as well... Anyways, if no one has done any profiling here, I would like to at least do that so we have some actual data as to where the bottleneck is. My test Wiki's are not large enough to give good data on this. What is the preferred way to go about testing this? Can someone who has access to the wiki server do the profiling for me? Can I get access to it to do it myself? Or can a copy of all the wikidata be made available to me so that I can test on my own systems? Ray
