Friday September 24, 2010

[Time] NameMessage
[01:21] jhawk28 does any other protocols than tcp work on Windows?
[11:27] Zao Assuming you manage to build it, I can't see why OpenPGM wouldn't work.
[13:45] CIA-20 jzmq: 03Stefan Majer 07master * r7073557 10/ src/org/zeromq/ : Javadoc, reduction of the visibility of a lot of flags reduced to private because they are only used inside public setters/getters. Fix of some checkstyle warnings. -
[13:45] CIA-20 jzmq: 03Stefan Majer 07master * r34d656f 10/ src/org/zeromq/ :
[13:45] CIA-20 jzmq: Deprecation of ZMQ.UPSTREAM and ZMQ.DOWNSTREAM
[13:45] CIA-20 jzmq: Socket.subscribe and Socket.unsubscribe only on a Socket of Type ZMQ.SUB -
[13:45] CIA-20 jzmq: 03Stefan Majer 07master * r86512f8 10/ src/org/zeromq/ :
[13:45] CIA-20 jzmq: Revert the subscribe/unsubscribe check,
[13:45] CIA-20 jzmq: Whitespace and identation with zeromq rules -
[13:45] CIA-20 jzmq: 03Gonzalo Diethelm 07master * re4abb79 10/ src/org/zeromq/ : Merge branch 'master' of -
[13:45] CIA-20 jzmq: 03Gonzalo Diethelm 07master * r3ba6cc1 10/ src/org/zeromq/ : Pulled changes from Stefan Majer to improve documentation and visibility. Made minor modifications. -
[14:53] codebeaker hey, quick sanity check - can I (should I?) include the zeromq repository in my own project as a git submodule ?
[15:16] cremes codebeaker: that's a good question for some git experts... i don' t think there are any in this channel
[15:18] codebeaker cremes: I don't need a git expert, my question was more about whether it makes sense, if you distribute a closed source application, between a couple of devlopers (mac and windows targets) how are other people typically including the ømq sources in their projects
[15:18] codebeaker I noticed there were separate downloads available for Windows and Unix, so that would lead me to believe you have to build them differently
[15:20] cremes i don't know how to answer your question... hopefully someone else will come along who can
[15:21] codebeaker thanks cremes - appreciate the attention anyway
[15:21] codebeaker basically, I could include the whole repository from ømq in my own, and then specify build flags - but I'm not sure how smart that would be
[15:31] pieter_hintjens codebeaker: including 0MQ as a git submodule would be a pretty good solution, yes
[15:32] pieter_hintjens at least until 0MQ becomes a standard package on the various distros
[15:32] pieter_hintjens and definitely a good idea if you want specific versions or branches of 0MQ
[15:33] pieter_hintjens but the distribution packages are identical except for line endings and compression format
[15:35] codebeaker pieter_hintjens: ^ thanks that's *exactly* the info I needed :-D
[15:35] codebeaker bew, ømq is freakin awesome
[15:36] codebeaker btw*
[15:36] pieter_hintjens codebeaker: hopefully it's also accurate :-)
[16:12] mjw9100 folks, just confirming that an "ordinary" fd for zmq_poll() can be an epoll fd?
[16:19] mjw9100 i.e. can I set zmq_pollitem_t.fd to an epoll() fd?
[16:19] mjw9100 according to the epoll() Q&A, these epoll fd's are selectable, etc.
[16:24] jhawk28 hello
[16:24] jhawk28 having a problem moving the compiled version of the windows libraries to another machine
[16:29] jhawk28 it looks like it is related to the following link:
[16:55] praveenbm5 hello all
[16:56] praveenbm5 i hope i am in the right room to discuss ZeroMQ
[16:56] codebeaker praveenbm5: the name says it all!
[16:56] praveenbm5 is anybody alive here
[16:57] praveenbm5 i was looking on the website regarding some guidance using zmq_poll
[16:57] praveenbm5 i just posted a comment there
[16:57] praveenbm5
[17:03] codebeaker ^ I'm inqo ømq for about 5 minutes now, what's the behaviour if you broadcast a message to an `upstream` (is the terminology the right way round there ) - but nothing is listening there yet, is it lost ?
[17:04] praveenbm5 its not lost
[17:05] praveenbm5 queues exist even when the end point doesnt yet exist
[17:05] praveenbm5 i suggest you read the guide on
[17:05] codebeaker ah, ok - great - i'm actually looking into it for writing a plugin for winamp that fires off a list of whats playing
[17:06] codebeaker (and I can't be sure the `other` app is running)
[17:06] praveenbm5 sorry i got your questiong wrong
[17:07] praveenbm5 you can connect to an end point from a zmq socket before the endpoint exists
[17:07] jhawk28 codebeaker: the new terminology for "upstream" is "push"
[17:07] praveenbm5 but regarding messages
[17:07] codebeaker no problem, winamp loads .dlls like plugins, and I wanted that plugin to shout out (over zmq) "hey, something played in winamp" - what happens if something isn't listening ?
[17:08] praveenbm5 push is for parallel worker based execution
[17:08] jhawk28 it sits in the local queue till the socket can connect I believe
[17:08] jhawk28 upstream/downstream == push/pull
[17:09] jhawk28 the only one that doesn't "keep" the message is the pub socket when noone has connected (I believe)
[17:10] codebeaker jhawk28: great, and that local queue is transient, in my example - if the closed winamp, and the plugin unloaded - it would be lost, right ?
[17:12] jhawk28 yes, I think so
[17:12] codebeaker jhawk28: thanks, I'm going to screw around in the Ruby console a little - but that sounds like exactly the behaviour I was looking for
[17:12] jhawk28 you would have to manually persist if you dont want it transient
[17:13] codebeaker yeah, jhawk28 - I'm looking for it not to persist, or at least to know if it was read off (I suppose I could always examine the local queue)
[17:13] jhawk28 easiest way to see is to download one of the samples and then compile it
[17:13] jhawk28 start them both a couple of times, but in different orders
[17:13] codebeaker also, lsat question (and I can infer a litlte from the guide…) the ipc:// dsn specification, it is portable across windows/unix, I assume ?
[17:13] jhawk28 I havent gotten ipc to work on my windows machine
[17:13] jhawk28 the inproc seems to
[17:14] codebeaker ahh - I was sortof counting on that - however since it's trivial, and it's not *really* IPC, if it's two separate apps… a .dll in winamp, and something else standalone - maybe IPC is the wrong paradigm now
[17:14] jhawk28 from the man page "The inter-process transport passes messages between local processes using a system-dependent IPC mechanism."
[17:15] jhawk28 vs "The in-process transport passes messages via memory directly between threads sharing a single ØMQ context"
[17:15] codebeaker ah, so IPC is certainly somehow better suited, but a socket would work just as well
[17:19] jhawk28 they are both "sockets", but the IPC is dependent on the OS
[17:19] codebeaker ah, ok
[17:19] jhawk28 so if you want cross platform, you probably need inproc
[17:19] guido_g no
[17:19] codebeaker inproc wouldn't work for talking from a winamp process to another process all together
[17:19] guido_g ipc is between different processes
[17:20] guido_g on windows, where ipc is not available, use tcp://
[17:20] codebeaker I assume I need to go out-of-band if IPC doesn't work on Windows, and use a `real` socket, like TCP or sth. that windows supports
[17:20] guido_g ipc is a real socket
[17:20] codebeaker ah, but a unix file socket, right ?
[17:20] guido_g right
[17:20] codebeaker ok, I assumed (foolishly) that ipc:// on windows would fall-back to tcp://
[17:22] jhawk28 so, inproc is between threads, but not between processes
[17:22] jhawk28 ?
[17:22] codebeaker jhawk28: certainly sounds like that
[17:27] jhawk28 has anyone done the zmq and jzmq build on vs2010 on Win7_64bit?
[17:28] guido_g have a look at the mailing list, it's been discussed there afair
[17:53] jond codebreaker: ipc isnt implemented on windows, inproc is between threads.
[18:01] codebeaker jond: rgr that (from earlier!) - pretty sure i'll go with a tcp socket on both platforms
[18:38] jhawk28 hurrah! I got VS2010 to compile zmq and jzmq
[18:38] codebeaker w00t, well done :-D
[19:18] rgl hi
[19:23] codebeaker hi rgl
[20:18] jond codebeaker: ipc is faster on linux, than tcp loopback; on solaris it made no difference as solaris loopback is optimised
[20:19] jond sustrik: how did the presentation go?