Wednesday September 7, 2011

[Time] NameMessage
[00:33] CIA-32 rbzmq: 03Michael Granger 07master * r3bc2e8b 10/ rbzmq.c : Use ZMQ::Error instead of RuntimeErrors to make exception-handling around sockets more predictable. -
[00:33] CIA-32 rbzmq: 03Michael Granger 07master * r1755fe7 10/ rbzmq.c : Whitespace fix -
[00:33] CIA-32 rbzmq: 03Brian Buchanan 07master * r8473116 10/ rbzmq.c : Merge pull request #13 from ged/distinct_exception_class ...
[00:35] CIA-32 rbzmq: 03Brian Buchanan 07master * r6c533c1 10/ extconf.rb : Merge pull request #11 from skaes/issue10 ...
[13:30] pieterh sustrik: ping
[13:35] mikko p to the h
[13:35] pieterh hi mikko :-)
[13:35] pieterh random question, have you ever used epgm://?
[13:35] mikko not really
[13:36] pieterh i'm experimenting with this for some p2p group messaging
[13:36] pieterh my idea is to hash group names to multicast addresses
[13:37] pieterh I guess I'll use the "poke a stick at it and see if it moves" technique
[13:44] drbobbeaty pieterh: I've actually used the hashing technique you talk about to broadcast messages over 270+ epgm:// channels. It works very nicely.
[14:44] sshirokov Is it safe to run 2.1.9 and 2.1.7 peers together or should I bring down everything, mass upgrade and bring back up on the same version?
[14:47] cremes sshirokov: i believe all of the 2.1.x series are mutually compatible
[14:47] cremes so you should be able to mix them
[14:48] sshirokov That doc'd anywhere?
[14:48] cremes no; why would you think that a patch release would be incompatible with an earlier patch release?
[14:48] sshirokov Just being overly cautious
[14:49] cremes ok
[15:10] pieterh drbobbeaty: hi
[15:11] pieterh sshirokov: yes, the interoperability is guaranteed by the version number
[15:11] pieterh
[15:11] sshirokov Thanks!
[15:12] pieterh drbobbeaty: stupid question perhaps, but does epgm have simpler requirements wrt routers?
[15:13] drbobbeaty pieterh: not really, but it's not bad. You just have to make sure the UDP multicast channels are available across the routers/switches. In our case, we found that we had some switches that only allowed 1000 UDP Multicast channels - this was a problem as we wanted to have many more than that. We have ended up getting new switches that can carry far more multicast channels as a result.
[15:14] pieterh aight
[15:14] drbobbeaty pieterh: In general, TCP is the most easily routed, but in a limited environment, the epgm is very nice.
[15:15] pieterh drbobbeaty: how does it work with more than 255 devices?
[15:16] drbobbeaty We've never pushed it that hard. Right now, I'm looking at about 20 receivers for a given sender - tops.
[15:37] crazed hey, i'm trying to install the perl bindings, but running into a strange issue
[15:37] crazed /usr/bin/ld: cannot find -luuid
[15:37] crazed during the make
[15:56] mikko crazed: cant link with lib uuid
[16:02] dcolish will there be a 2.1.10 or will 2.1.9 -> 2.2?
[16:11] pieterh dcolish:
[16:11] dcolish thanks
[16:11] pieterh there will be a 2.1.10, yes
[16:11] pieterh there will be maintenance releases of 2.1.x as long as there are users and patches
[16:28] dcolish ok thats good to know, thanks pieterh
[16:29] dcolish it doesnt look like issue-207 would help my problems, but i'm much more aware of how name resolution works for tcp:// in zmq
[16:39] pieterh dcolish: what are your problems?
[16:47] dcolish well, we forgot to roll out a dns host name and that causes the zmq resolver to fail
[16:48] dcolish i dont think its zmq's responsibility to fix that
[16:48] dcolish in fact, its really nice that it crashed early
[16:49] mikko pieterh: sent you an email
[16:49] mikko this is like live notification system
[18:16] pieterh hi mikko, was out at lunch
[18:33] mikko pieterh_: cool