<div>On 11/3/05, seth vidal <<a href="mailto:skvidal@phy.duke.edu">skvidal@phy.duke.edu</a>> wrote:<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"> you think async downloads will actually be all that much of an improvement? As a test setup entirely local-disk repositories and run a chroot install, compare that to a fast network and see where the bottlenecks are. I'm curious of the results. The biggest objection I've had to async downloads/parsing is that you end up having a very confusing callback interface to track what is being downloaded and when. Your thoughts on correcting that? </blockquote></div> Never said it would be easy, I only said it would be wor....@#$(@*.. Never mind. Anyway, as to the first question: Well buildgroups needs to be pulled down locally. (For me anyway). I "feel" like the setup stage is taking quite awhile even with the --no-clean param. But, I haven't done any legit tests or looking into why. I presume it's the buildgroups thing, but, maybe something else lurks. Async definitely needs callbacks, there's no other way. Async and threaded programming never come through cleanly, unless you want to use Twisted. But, that'd be a bulky dep to have for yum. :P Asyncore is in Py; but, many heads have declared it crap. Not sure what smart is using, because I haven't pealed it open yet. But, smart is definitely downloading async or multi-thread. It's definitely an eye-opener difference. Of course, this diverges from the topic about local buildsystems. So, I'll shut up now. thanks, jeff -- -jeff