Sunday March 13, 2011

[Time] NameMessage
[00:04] nadime i am using inproc for anything local, cremes
[00:04] nadime the only thing i use tcp for is external
[00:05] nadime thanks for the suggestions, re: instrumentation
[00:06] nadime oh incidentally, mikko, we were doing the math wrong. 50B x 500k = 25MB/s
[00:06] nadime which makes much more sense
[00:06] nadime since i'm seeing 10% utilization
[00:06] nadime 10-15
[01:03] Guest51726 hello, anyone awake tonight ?
[01:04] Evet :)
[01:17] Guest51726 Hello Evet, how are you ?
[01:17] Evet fine, Guest51726. you?
[01:18] Guest51726 not so bad !
[01:19] Guest51726 I was doing some experimenting with 0MQ, 2.1.2 the other day - tell me, is the load balancing stable ?
[01:26] Guest51726 Yep.
[01:41] Evet Guest51726: i think so
[01:41] Guest51726 hmmm...that wasn't what I was seeing :(
[02:04] emacsen I have a, hopefully basic question: I'm learning zeromq. I only have my laptop and I'm trying to test push and pull requests. I heard that pull requests will be balanced across pullers. But I'm testing with ipc and all the requests go to one listener. Is that because of ipc?
[02:05] Guest51726 thats a good question: I don't know.
[02:05] Guest51726 the requests really should be balanced across pullers, regardless of the transport. However I have observed uneven load balancing myself.
[06:19] guido_g hello world!
[06:34] pieterh hey guido_g, hi!
[06:34] guido_g morning pieterh
[06:35] pieterh another day, another little protocol spec :-)
[06:35] guido_g just pulled the zeromq2-1 changes and compiled
[06:35] guido_g ohh...
[06:35] guido_g spec:10
[06:36] pieterh yeah, brokerless reliability
[06:36] pieterh final part of chapter 4... finally... :-)
[06:36] guido_g yes!
[06:36] guido_g :)
[06:37] pieterh for 2.1 I defined ZMQ_DEALER and ZMQ_ROUTER as aliases
[06:37] guido_g seen it
[06:37] guido_g good move imho
[06:38] pieterh I'm going to change the man pages, leave the old names as deprecated
[06:38] pieterh this will make it easier for people to move to 3.0
[06:38] guido_g ok
[06:39] guido_g from spec10: "From the ØMQ Reference Manual[Bibliography item zmqsocket not found.]:"
[06:39] guido_g seems that a link can't be resolved
[06:39] pieterh just fixed that, thanks
[06:39] guido_g ahh
[06:40] guido_g reload did it
[06:40] pieterh I figured out how to do router-to-router connections between clients and servers
[06:40] guido_g great
[06:40] guido_g will be fun to read
[06:40] pieterh it's rather fun because neither side can talk to the other without knowing their identity first
[06:41] guido_g using ahh... router sockets
[06:41] guido_g need to get used to the new name
[06:42] pieterh :-)
[06:43] guido_g example program adjusted
[10:47] Guthur pieterh, Ch4 is turning into a bit of a monster
[10:48] Guthur good stuff though
[10:49] Guthur After sorting out my POSIX issues, and fix a bug I found during the process, I'm now ready to finish ch3
[10:49] Guthur looking forward to ch4
[16:02] pieterh Guthur: CH4 is now finished
[16:03] Guthur cool, i'm just debugging my peering3 and then i can start on ch4
[16:03] pieterh :-)
[16:17] mikko \\o
[16:17] mikko o//
[16:18] mikko pieterh: we got some more fixes going on in openpgm with steve
[16:18] pieterh mikko: nice, what's the issue?
[16:18] mikko sun linker visibility with sungcc compiler
[16:18] pieterh ah, ok
[16:18] mikko and i think i've found zeromq compile warning on freebsd 8.1
[16:19] mikko pgm_receiver.cpp:154: warning: dereferencing type-punned pointer will break strict-aliasing rules
[16:19] pieterh that definitely looks like a warning :-)
[16:19] mikko it breaks the build nowadays
[16:19] mikko previously zeromq was compiled without -Werror when openpgm was enabled
[16:19] mikko now that openpgm is built separately i added -Werror back in
[16:23] pieterh mikko: sanity check, do you see a man page for zmq_device() in your zeromq source tree?
[16:24] mikko
[16:24] mikko doesn't seem to be there
[16:24] pieterh funny, the method isn't documented, but I recall writing a man page for it
[16:25] pieterh which is installed on my box...
[16:25] pieterh hmm
[16:25] mikko maybe you haven't committed / pushed it?
[16:25] pieterh nope, it was reverted by the maintainers...
[16:25] pieterh sep 4
[16:26] pieterh ok, I'm adding it back to 2.0 and 2.1
[16:26] mikko when are you flying back?
[16:27] pieterh as soon as these mushrooms wear off
[16:27] pieterh what? what?
[16:27] pieterh :-) I'm not away, Mikko, happily working at home
[16:27] mikko i mean from london
[16:27] pieterh ah, I take the train, easy ride from London to Brussels
[16:27] mikko ah, ok
[16:27] pieterh if you want to meet up I'm around Wednesday afternoon
[16:27] mikko i noticed that cloudcamp ( is on 17th
[16:28] mikko if that is of interest to you
[16:28] pieterh sigh, yeah, won't make it
[16:28] mikko i'm gonna be on client site all next week so probably earliest we meet is in the pub
[16:28] mikko been hectic last week
[16:28] mikko setting up my own company
[16:29] pieterh nice!
[16:29] mikko almost everything ready
[16:29] mikko still need to sort out bank account
[16:29] mikko (i think at some point it would be nice to get paid)
[16:31] pieterh that's always a nice option
[17:44] Guthur I'm getting an "Operation cannot be accomplished in current state" in the clientTask of peering3
[17:45] Guthur it works for a while then that error on a recv
[17:45] Guthur any suggestions on how to find the route of the error
[17:53] mikko Guthur: req/rep sockets?
[17:53] pieterh hmm, sounds like a REQ socket state error
[17:53] pieterh let me check that
[17:53] mikko yeah, was thinking the same
[17:53] Guthur yeah re
[17:53] mikko wrong order of recv/send causes that
[17:53] Guthur req/rep
[17:53] mikko usually
[17:54] mikko so for example send/recv/recv on req socket would cause that
[17:54] Guthur oh ok, it's possible there is an error with my code
[17:54] pieterh Guthur: it could be a bug in the design, as well
[17:54] mikko there possibly should be a way to recover the socket
[17:55] pieterh well, what I do is close it and reopen, in later examples
[17:55] pieterh kind of annoying but finally it works ok
[17:59] Guthur dinner time, I'll check my code later
[19:10] guido_g <- zeds ømq talk at pycon
[19:22] mikko interesting observations
[19:25] Skaag mikko: getting paid is very important, but can be dangerous if not done right
[19:25] Skaag need to consider it strategically
[19:26] Skaag but no doubt, it has to happen
[19:32] mikko Skaag: i really haven't got much savings so there isnt that much room for strategy
[19:32] mikko :)
[19:32] Skaag well ok but here's what I mean:
[19:33] Skaag Assuming you have a client who wishes to hire your services - what does he need exactly?
[19:33] mikko services?
[19:33] Skaag And the question is, of course, how many such offers you get
[19:33] Skaag the danger i'm talking about is one of stifling innovation
[19:33] Skaag basically the elimination of time for creativity
[19:34] Skaag sometimes you need to delegate work to a team, make a cut from the top, but remain "CTO" and continue to research and develop the core technology
[19:34] Skaag you don't have a day job?
[19:34] Skaag I was certain you were hired for good pay at the very least :)
[19:34] mikko i have but not soon
[19:34] Skaag company closing?
[19:35] mikko i resigned to start my own company
[19:35] Skaag ah, good luck!!
[19:35] mikko thanks
[19:35] Skaag it's not easy, the path you have chosen
[19:35] Skaag :)
[19:35] Skaag but certainly, most fulfilling :)
[19:36] Skaag some would say, easiest is to be the cow in the shed, slowly milked, given music and food...
[19:36] Skaag :)
[19:36] mikko probably
[19:36] mikko i like the excitement of setting up a company
[19:37] mikko even if it fails eventually
[19:37] mikko knowing that there won't be food on the table unless you do something about it
[19:38] guido_g hmmm i gained ~20kg in the 2 years since i went freelancing
[19:38] mikko sustrik: there?
[19:43] Skaag mikko: I so know what you're talking about, that I can't help but remember all the pain :)
[19:44] Skaag If you need any help, I am here
[19:44] mikko i'm not overly worried at the moment
[19:44] mikko not yet at least :)
[19:52] Skaag what's the goal of the company?
[19:53] mikko i got a few SaaS ideas but i guess initially just bring the food to table
[19:53] mikko initially providing services i would guess
[19:54] Skaag ok so the classic "Tech Guy opens a Company" scenario, as I thought :)
[19:54] mikko yeah
[19:55] mikko initially it will just be me
[19:55] Skaag (not saying it in any negative way!)
[19:55] mikko i haven't really planned things very far
[19:55] mikko i just kinda felt that i need new challenges
[19:55] Skaag I did it for other reasons
[19:55] Skaag I wanted a stable passive income to allow me to research things in my free time
[19:56] mikko did it work out?
[19:56] Skaag corporate took too much of my time, they were the hungry pig that never had enough
[19:56] Skaag actually yes :)
[19:56] Skaag thanks to my wife though
[19:56] Skaag if she hadn't taken the financial parts from me, it would have failed miserably
[19:56] Skaag not ashamed to admit it
[19:57] Skaag she's a woman who is not afraid to bill customers for her husband's work/hours :)
[19:57] Skaag and she will hunt them down, until hell freezes, or they pay!!
[19:57] Skaag and better not tell her the payment is late, for she will unleash upon you her furious anger, and terrible might!! :)
[19:58] Skaag ok I may be exaggerating it a bit, but more or less that's how it is! :)
[19:58] mikko in UK it seems very common to delay 50+ or so days
[19:58] Skaag never go down that route
[19:58] Skaag never.
[19:58] Skaag do not agree to that.
[19:59] mikko no, i mean companies just not paying even when the terms say 30 days or so
[19:59] Skaag don't even agree for 30 days
[20:00] mikko 30 days is common here
[20:00] Skaag 7 days from the presentation of the invoice.
[20:00] mikko back in finland 7 days was reasonable
[20:00] mikko here, i don't think so
[20:00] Skaag and I don't know how tax authorities work in the UK, but VAT in Israel want their chunk of the money whether or not you received payment, for any invoices you produced.
[20:01] Skaag mikko, seriously, who cares what is common
[20:01] Skaag start with 7 days and let the customer squirm
[20:01] Skaag and then if you do heed the 30 days, at least he will feel he won something
[20:01] Skaag he earned them ;)
[20:01] Skaag I hate seeing people get fucked, especially if I care about them and their future
[20:04] mikko pieterh: have you done work on zns or still gathering ideas?
[20:06] nadime if mikko or cremes are here -- solved my IO bottleneck problem
[20:06] nadime profiler did the trick
[20:14] michelp i have to agree with Skaag on this one, when the customer balks that 30 days is normal, tell them to go hire a normal person. you're better than normal.
[20:18] Skaag exactly my point - this is Mikko!!!! Not some random developer.
[20:18] Skaag I mean c'mon, is nothing holy anymore?!
[20:18] Skaag :)
[20:38] mikko nadime: what was the issue?
[22:17] nadime mikko - i was using a poller in a bottleneck thread for no good reason
[22:40] Guthur should a SUB socket connect block until it's connected to an endpoint?
[22:43] Guthur the example in peering3 seems to block
[22:44] Guthur but mine does not which has me scratching my head
[22:44] yo yo
[22:44] yo what's up