[04:05] *** salmon_ has joined #kamaelia |
[06:11] *** Uraeus has joined #kamaelia |
[12:02] *** Uraeus has joined #kamaelia |
[15:17] *** MS- has joined #kamaelia |
[15:17] < MS-> evening |
[16:12] < Lawouach> evening |
[16:12] *** vmlemon_ has joined #kamaelia |
[16:12] < Lawouach> http://developers.facebook.com/news.php?blog=1&story=301 wondering how K. would have performed |
[16:19] < vmlemon_> Hi |
[16:22] < Lawouach> hi vmlemon_ |
[16:22] *** vmlemon_ was surprised to see that code released so soon, after the acquisition of FriendFeed |
[16:48] *** vmlemon__ has joined #kamaelia |
[16:49] *** vmlemon__ is now known as vmlemon_ |
[16:59] *** eikenberry_ has joined #kamaelia |
[17:12] < MS-> I think Kamaelia would've performed awful on that test for what it's worth, for a couple of reasons. |
[17:12] < MS-> Firstly, we've not really done the optimisations that would make that test fly |
[17:12] < MS-> secondly, we use select, not epoll |
[17:12] < MS-> thirdly, I'm not convinced of the performance of the HTTP code |
[17:13] < MS-> That said, I don't think the performance would be bad really - awful in terms of comparison, but not bad |
[17:14] < MS-> Also, I think when I get a chance to work on axon2 properly for merging that would change things quite dramatically since that is a significant performance difference |
[17:14] < MS-> However without _trying_ these things it's pure speculation. |
[17:14] *** salmon_ has joined #kamaelia |
[17:15] < MS-> I think though, I'm not too worried about that tbh |
[17:15] < MS-> My aim really was always for the system to be _optimisable_ to something very fast, rather than to build something optimised up front |
[17:16] < MS-> But then that goes with my general ethos of "get something working; check you're doing the right thing; make it fast" |
[17:42] *** eikenberry_ is now known as eikenberry |
[18:05] *** vmlemon_ wonders if it'd integrate cleanly as a Kamaelia component |
[18:06] < vmlemon_> *encapsulated as a |
[18:06] < vmlemon_> *within a |
[18:06] < vmlemon_> Meh |
[19:05] < MS-> Oh, yes, you'd be able to make their code a component. |