Saturday, 2014-03-08

*** tpb has joined #pyconau00:00
*** tserong has joined #pyconau00:41
*** 16WAA1JKJ has joined #pyconau01:28
*** 16WAA1JKJ is now known as jhesketh101:29
*** jhesketh1 has quit IRC02:32
*** tserong has quit IRC02:41
*** tserong has joined #pyconau02:41
*** tserong has joined #pyconau02:41
*** jhesketh1 has joined #pyconau03:25
jammyjamprologic: so when I look at the page, my first thought is "sounds like a similar problem space to twisted" - is that on the right track?04:32
jammyjamprologic: fwiw, I don't know about the pyconau papers committee, but I know that the LCA papers committee often turns away talks that are good and interesting that they'd like to see, but couldn't find space for. I *think* pycon-au also gets enough talks that they have to turn away good talks, so I wouldn't interpret not making the conf last year as a sign that it wasn't interesting04:33
jammyjamprologic: I would hope that the papers committee from last year would be able to give you feedback *cough* chrisjrn *cough*04:34
prologic<jammyjam> prologic: so when I look at the page, my first thought is "sounds like a similar problem space to twisted" - is that on the right track? <-- in short - yes .. similar problem space. i.e: concurrent systems, async i/o04:34
prologicbut ofc Twisted and circuits share almost nothing alike as circuits was initially developd over 10 years ago04:34
prologicalso jammyjam appreciate your comments04:35
prologicfrom what you've said (and a few others) I think I will submit :)04:35
jammyjamWhat's the latest version of python that twisted supports? My memory tells me it's something like 2.5, but I could be wrong04:36
chrisjrnPIng me in a few hours to talk about this04:44
hawkowljammyjam, er05:05
hawkowljammyjam, 2.7 + pypy + portions on 3.205:05
hawkowljammyjam, (you're talking to the release manager of the latest Twisted release, 13.2 :D)05:05
hawkowlTwisted has test coverage on python 2.6, python 2.7, pypy and portions of py3 (mostly only twisted core) on windows, mac, debian, ubuntu and fedora05:06
jammyjamhawkowl: it's clearly much more up-to-date than my memories of it are05:07
hawkowljammyjam, it's still going, yes :D05:08
hawkowlin fact, good performance on pypy is a target, even05:08
hawkowlpypy actually use twisted in their speed tests05:08
hawkowl(there is a significant pypy-twisted developer overlap)05:08
*** jammyjam is now known as tchaypo05:18
*** jhesketh1 has quit IRC06:28
prologiccircuits supports 2.6, 2.7, 3.2, 3.3 and pypy :)07:39
prologicWindows ofc is a bit crappy - only select, crappy forking (non-existent?), etc07:39
prologic <-- just today Ryan Barber created this nice RPi clock with circuits ;) very impressive07:41
tpbTitle: rasp-clock (at
hawkowlprologic, twisted uses IOCP07:43
hawkowlprologic, which is crappy in itself (oh god having to write all your underlying TCP/UDP twice)07:43
hawkowli should probably put in a proposal about twisted stuff, I dunno07:44
*** prologic_ has joined #pyconau08:51
*** prologic has quit IRC08:56
*** prologic_ is now known as prologic08:56
prologichawkowl, how does Twisted have to redefine tcp/ip stuff twice?12:09
prologicI've not really looked too deeply at it's core12:09
prologicbut I did write a twisteedintegration for circuits allowing twisted apps and protocols to run on top of circuits (
tpb<> (at
jayveeprologic: the fact that Twisted *still* does not support IPv6 after all these years (something that even someone as dumb as I can comprehend perfectly) does not lend towards its credibility12:17
prologicwow you for real?12:18
*** kulich has quit IRC13:42
hawkowljayvee, er, no, it does support IPv614:13
tpb<> (at
hawkowlprologic, certain things need to be done in the IOCP reactor + in the regular reactors14:14
hawkowlprologic, anything using sockets is slightly different enough that you have to keep it separate14:15
hawkowldifferent flags, that sort of thing14:15
hawkowlthe tests are written once, it's just the reactor implementation which is a lot of subtly different because lol Microsoft code14:16
hawkowli'm sure it could be done better, but the reactor isn't really changed a lot14:19
*** prologic has quit IRC14:37
*** prologic has joined #pyconau14:44
*** prologic is now known as Guest4748314:44
*** Guest47483 has quit IRC15:58
*** Guest99731 has joined #pyconau16:01
*** Guest99731 has quit IRC16:02
*** Guest99731 has joined #pyconau16:02
tchaypohawkowl, prologic - what I really want to see now is a combined talk from the two of you16:57
tchaypotalking about the broader problem space and comparing the two approaches16:57
hawkowlwell, Twisted does much more than that problem space, to be honest16:57
tchaypofight! fight!16:58
hawkowltwisted has things like AMP, twisted.mail, twisted.conch, twisted.names...16:58
hawkowlTwisted soonish will have the infrastructure for a DNSSEC-capable DNS server16:59
hawkowl(have I mentioned you can write a mail-enabled application in like 40 lines of code, too)16:59
tchaypoYou make it sound like emacs17:04
hawkowlbut with less lisp one hopes17:06
tchaypoHas anyone written a lisp for the python VM?17:07
tpbTitle: Welcome to Hy’s documentation! hy 0.9.12 documentation (at
hawkowl"Since Hy transforms its Lisp code into the Python Abstract Syntax Tree, you have the whole beautiful world of Python at your fingertips, in Lisp form!"17:07
hawkowltchaypo, if you wanted, you could probably write a lisp interpreter in rpython and use the pypy vm17:09
hawkowlwhich is what Topaz is (ruby on the pypy vm)17:09
hawkowlor, rather, same underlying JIT as pypy17:09
tchaypoI was imagining that's how it would be done. Manipulating the AST directly seems so much harder17:10
hawkowlwell, it doesnt manipulate the ast17:10
hawkowlit writes ast17:10
hawkowlit is essentially a transcoder17:10
hawkowlbut a slightly roundabout one17:11
*** hzopak has joined #pyconau18:13
*** maksimmar has joined #pyconau21:37
*** jhesketh1 has joined #pyconau22:43
Guest99731morning guys23:28
Guest99731so yeah circuits's so-called reactors -- we call them pollers23:28
Guest99731are exactly that, just separate components, writen once23:29
Guest99731recently (last few years, releases maybe, can't exactly remember) we borrowed quite heavily from twisted reactor code to handle better many edge cases with various platforms23:29
Guest99731but this is not the important bits of circuits -- they're just I/O components like any other component23:30
Guest99731and yes twisted has twisted.mail, twisted.names, twisted.coudh, and lots of nice shiny protocol -- circuits only has what current and past users have used. irc, http (client and serer), lp (line protocol), websockets and not much else at this stage (we want more, but no-one's contributed/asked for more!)23:31
Guest99731that being said (as I mentioned above) circuits for a year or so now can actually reuse twisted protocols as-is as well as run most twisted apps on the circuits core23:32
Guest99731oh crap my nick :/23:32
*** Guest99731 is now known as prologic23:32
prologicsorry :)23:33
prologicstupid FreeNode network23:33
prologicseems unreliable these days, lots of DDoS :/23:33
Zimskymany ddos, much packet, very flood23:39
prologicyes that's one way of putting it23:39
prologicit's gotten so bad that I'm planning on adding distributed support to my logging bot that logs several chanells around here23:39
jayveemultipath IRC23:44
prologicyeah precisely23:48

Generated by 2.12.1 by Marius Gedminas - find it at!