Wednesday June 29, 2011

[Time] NameMessage
[07:32] kventil Hi, just a short question: if i plan to build a pubsub "hub" system i definitly need openpgm, right?
[07:38] sustrik you can use a forwarder device in the middle instead
[07:42] kventil sustrik: but this device has to subscribe "manually" to all existing publishers, right?
[07:42] sustrik nope
[07:42] sustrik publishers can connect to the device themselves
[07:43] kventil ah okay
[07:44] kventil so i have some kind of N (publishers) to 1 (forwarder) to N (subscribers) architekture?
[07:47] sustrik yes
[07:47] kventil thanks :)
[08:10] alkemann what is a good page for overview of all socket types and listing differences between them please?
[08:28] sustrik alkemann: zmq_socket(3)
[08:30] alkemann i prefer tinyurl or as url compression, but i guess that will work in a pinch
[08:30] eintr i like turtles
[08:31] alkemann though probably only because I was already looking at the api
[14:10] bcg czmq and libzfl seem to overlap quite a bit, is there a reason there are two high level C libraries being developed in parallel?
[14:11] bcg I am starting a C zmq app and was thinking of using one of these libraries
[14:14] sustrik that's a question for pieter hintejns
[14:14] sustrik try asking on the mailing list
[14:17] bcg ah it appears libzfl now depends on czmq
[15:38] ninjaspounced join #archlinux
[16:30] Steve-o IOCP doesn't appear too difficult to integrate into 0mq
[17:56] nicolas is it correct to use multiple publishers to a single sink subscriber?
[18:06] ptrb I've got a long-running process that ZMQ_SUBs to a ton of data. It's growing in memory reliably, and valgrind reports the vast majority in "still reachable" blocks -
[18:06] ptrb nicolas_: it works, if that's what you mean.
[18:13] ptrb I guess my question re: above -- is it keeping this memory wired deliberately? if so, is there some way to signal it should be released?
[18:26] cremes ptrb: are you sure the subscriber is all caught up?
[18:26] ptrb cremes: I see. No, I'm not.
[18:26] cremes then that memory is holding queued up messages (likely)
[18:26] ptrb Interesting. Thanks for the insight.
[18:27] cremes is that surprising to you?
[18:27] ptrb No, it makes a lot of sense. I just didn't think about that being a possible culprit :)
[18:27] cremes oh, ok
[18:27] cremes if your publishers are faster than your subscribers, you'll eventually exhaust memory
[18:27] cremes i think this was the original reason for adding HWM