Three points: It seems that using "-C" speeds things up *a lot* on all of my systems, which are all using "mirrorlist" (as is the default in FC3.) From discussion here, I gather that that means they're going to hit the net each time, and get a randomly-different server each time. I'd suggest that the current mirrorlist behavior should be changed to be more stateful: yum should remember the last mirror that it used, and just stick with that for a while. In other words, always behave as if -C was specified unless: A) it's been more than N hours; or B) an error occurred trying to download something from the currently-selected mirror. Second: everyone keeps talking about dependency resolution and how that's bound to be slow, but as far as I can tell, there's no reason to resolve dependencies at all when I'm just doing "list". So either there's dependency resolution going on when there shouldn't be, or it's not relevant at all. Third: Konstantin Ryabitsev wrote: > > I do not believe this is where we should look for to speed up the > performance of the package management system. The slowness is not in > yum, it's in the systems it relies on to function. I think the theory that "yum is fine, it's the RPM libs that are slow" is pretty well debunked by this: % time rpm -qa '*mtr*' >/dev/null 3.576u 0.121s 0:05.93 62.2% 0+0k 0+0io 0pf+0w % time yum -C list '*mtr*' >/dev/null 5.474u 0.539s 0:16.41 36.5% 0+0k 0+0io 8pf+0w I don't think "three times longer" is exactly negligible. -- Jamie Zawinski jwz@xxxxxxx http://www.jwz.org/ jwz@xxxxxxxxxxxxx http://www.dnalounge.com/ http://jwz.livejournal.com/