ZeroMq IRC Log

Wednesday August 4, 2010

[Time] Name	Message
[02:02] ateTate	From experiementing w/ zeromq today, it seems like if a request socket does not successfully recieve after sending, the message is not sent. Why is that?
[12:45] CIA-19	zeromq2: 03Pieter Hintjens 07master * r11a410b 10/ (11 files in 2 dirs):
[12:45] CIA-19	zeromq2: Renamed ZMQ_UPSTREAM to ZMQ_PULL, and ZMQ_DOWNSTREAM to ZMQ_PUSH. Left the old
[12:45] CIA-19	zeromq2: definitions as aliases, to be removed in release 3.0. Also renamed the source
[12:45] CIA-19	zeromq2: files implementing these two socket types. This change does not break existing
[12:45] CIA-19	zeromq2: applications nor bindings, but allows us to fix the documentation and user guide
[12:45] CIA-19	zeromq2: now, rather than keeping the old (confusing) names. - http://bit.ly/bXUNom
[12:45] CIA-19	zeromq2: 03Pieter Hintjens 07master * rf575f25 10/ doc/zmq_device.txt : Added man page for the zmq_device method - http://bit.ly/cOwku4
[12:45] CIA-19	zeromq2: 03Pieter Hintjens 07master * r6cd9030 10/ doc/Makefile.am : Added clean target that deletes generated man pages - http://bit.ly/9jsaSI
[12:45] CIA-19	zeromq2: 03Pieter Hintjens 07master * r77a3c36 10/ (33 files):
[12:45] CIA-19	zeromq2: Various changes to documentation project:
[12:45] CIA-19	zeromq2: * Added documentation for zmq_deviced, which we're developing
[12:45] CIA-19	zeromq2: * Created consistent page footer in documentation template
[12:45] CIA-19	zeromq2: * Page footer notes doc authors and copyright statement - http://bit.ly/d2130a
[12:57] CIA-19	zeromq2: 03Pieter Hintjens 07master * rc51de31 10/ doc/Makefile.am : Reverting 'clean' change to Makefile - http://bit.ly/93Hkm2
[13:07] CIA-19	zeromq2: 03Pieter Hintjens 07master * r6ff1939 10/ (4 files): Removed empty man pages for old standalone devices - http://bit.ly/csTExH
[13:27] mato	pieterh: ok, i'm here now
[13:28] pieterh	mato: any particular advantage in breaking manual text into lines?
[13:28] mato	pieterh: manual line breaks mean that a) the doc text can be meaningfully read with less in a terminal b) i can edit it with vim/emacs/other standard unixy editor
[13:28] mato	pieterh: also, the long lines don't show up on github nicely
[13:28] pieterh	ok, i
[13:28] mato	see e.g. http://github.com/zeromq/zeromq2/blob/master/doc/zmq_deviced.txt
[13:29] pieterh	i'll buy the github argument
[13:29] mato	hey, it's a .txt file... 80 columns is your friend :)
[13:29] pieterh	perhaps in 1992, yes
[13:29] pieterh	it's been a long long time since 80 columns had any mystical value
[13:29] guido_g	no, noone knows how wide you editor windows was
[13:29] pieterh	everyone knows today the magic number is 140
[13:30] mato	well, please just use 80 column text, thanks
[13:30] mato	how about an old one? vim? emacs? :)
[13:31] guido_g	:)
[13:31] mato	i'm sure even gedit or whatever it's called can format text :)
[13:35] pieterh	mato: would you review zmq_device(3) in trunk and confirm it makes sense?
[13:35] mato	pieterh: ok, but i'll have to reformat it :)
[13:36] pieterh	i'll do that, mato
[13:36] mato	pieterh: also, it's 'master' not trunk :)
[13:36] pieterh	am in process of doing so
[13:36] mato	thx
[13:36] pieterh	and yes, it's master, not trunk, my bad
[13:44] pieterh	uhm, mato
[13:44] mato	yes?
[13:44] pieterh	there are quite a few instances of lines longer than 80 chars in the text files
[13:44] pieterh	i'll fix them all up
[13:44] mato	there are places where due to syntax limitations in asciidoc you can't wrap lines properly
[13:44] mato	please wrap "normal text body" properly
[13:44] mato	thx
[13:45] pieterh	? asciidoc lets you do it properly
[13:45] pieterh	also its "its" not "it's" when you use the possessive
[13:45] pieterh	grammarnazi is on the loose
[13:45] mato	i think you are referring mainly to the SYNOPSIS section?
[13:45] pieterh	nope, random text
[13:46] mato	and unfortunately those don't render properly when wrapped
[13:46] pieterh	egrep ".{80,}" *.txt is your friend
[13:46] pieterh	i'll fix it all up and double check it renders properly when wrapped
[13:47] mato	what i see are mostly SYNOPSIS sections and definition lists
[13:47] mato	if you can get them to format properly when wrapped (check manpage result and HTML) then by all means fix them
[13:48] pieterh	let's give it a shot :-)
[13:48] pieterh	how do I generate the html?
[13:48] mato	make dist will do it
[13:49] pieterh	"no rule to make target dist" when I run that in /doc
[13:49] pieterh	np
[13:49] mato	make dist at toplevel
[13:49] pieterh	ack
[13:49] pieterh	man pages format correctly, will now check html
[13:50] pieterh	seems fine
[13:50] mato	hmm, can you commit it, i'll double-check here (since i build releases on this machine)
[13:50] mato	that's if you're done...
[13:51] pieterh	tried on some docs, am now doing rest and checking individually
[13:51] mato	also i assume you're not going to reformat everything just the wrong bits?
[13:51] pieterh	only the lines that are more than 80 chars, mato
[13:51] mato	thx
[13:51] pieterh	also some spelling mistakes as i find them
[13:52] pieterh	minor stuff
[13:52] mato	(a UK spell checker, to be precise)
[13:52] pieterh	its/it's in a few places
[13:53] mato	ah, sorry, i have a problem with that one for some reason
[13:53] pieterh	me too, i get it right but always think i got it wrong
[13:53] pieterh	just remember, his, hers, its
[13:53] pieterh	not hi's her's it's
[13:54] pieterh	and it's, they're, we're, you're
[13:54] pieterh	not its, theyre, were, youre
[13:54] pieterh	:-)
[13:54] pieterh	english can be remarkably consistent
[14:07] CIA-19	zeromq2: 03Pieter Hintjens 07master * r13f3481 10/ (10 files):
[14:07] CIA-19	zeromq2: Further cleanups on reference manual
[14:07] CIA-19	zeromq2: - fixed unwrapped text in new man pages
[14:07] CIA-19	zeromq2: - fixed over-long lines in older pages, where possible
[14:07] CIA-19	zeromq2: - removed reference to old standalong devices from index page
[14:07] CIA-19	zeromq2: - added refernce to new zmq_device[3] documented from index page
[14:07] CIA-19	zeromq2: - some minor spelling corrections - http://bit.ly/d9NCpE
[14:07] pieterh	mato: ok, done, double-checked, and committed
[14:07] pieterh	wow, CIA-19 got there before me :-)
[14:07] mato	:-)
[14:09] mato	-.Polling indefinitely for input events on both a 0MQ socket and a standard socket.
[14:09] mato	103
[14:09] mato	+.Polling indefinitely for input events on both a 0MQ socket and a TCP socket.
[14:09] mato	pieterh: perhaps pedantic, but 'socket' can be any old socket :)
[14:09] mato	no big deal
[14:09] mato	anyhow, thx
[14:10] mato	pieterh: which did you want reviewed? zmq_device?
[14:16] mato	pieterh: re zmq_device:
[14:17] mato	pieterh: you talk about "proxy" and "broker", but only provide examples for the 2nd one
[14:19] mato	pieterh: also, not sure a :: list is the best way to format the model list. I would have probably used either subsections, or XXX model. ...Body text...
[14:19] mato	pieterh: mostly I just don't just the lowercase "proxy" and "broker" :)
[14:20] mato	pieterh: also the references to "downstream" are a bit confusing, but i don't have a better idea
[14:20] mato	pieterh: that's all, thx
[14:55] pieterh	mato: re
[14:55] pieterh	zmq_device(3) covers the core of how devices work
[14:55] pieterh	zmq_deviced(1) covers the api for the command line tool
[14:57] pieterh	thanks for the review, i'll fix the text
[15:07] CIA-19	zeromq2: 03Pieter Hintjens 07master * r96bcc9e 10/ doc/zmq_device.txt :
[15:07] CIA-19	zeromq2: Small improvements to zmq_device(3) page
[15:07] CIA-19	zeromq2: * Clarified broker model and proxy model
[15:07] CIA-19	zeromq2: * Added example of proxy model - http://bit.ly/aH2jNL
[16:58] cremes	fyi, http://github.com/rabbitmq/rmq-0mq
[18:02] erickt	good morning! would it be possible to get dates displayed on the zeromq.org documents? I've found myself looking at the old 1.0 era docs and getting confused :)
[18:09] sustrik	erickt: please do complain on the mailing list, someone has to go through all the content and remove out-of-date pieces
[18:10] sustrik	cremes: yes, that's 0MQ/AMQP bridge project
[18:10] sustrik	see its wiki for design diagrams
[18:10] sustrik	however, implementation is still missing :)
[18:10] cremes	yes, i just wanted to highlight it for folks lurking in channel
[18:23] erickt	sustrik: sure
[18:24] erickt	i was just hoping that the wiki software had an option to turn on the dates :)
[18:35] pieterh	erickt: it does
[18:36] pieterh	i will need to modify some templates
[18:36] pieterh	give me a minute, I'll try this on the whitepapers section...
[18:37] erickt	pieterh: thank you so much
[18:45] pieterh	erickt: small catch
[18:45] pieterh	it works, see http://www.zeromq.org/docs:projects
[18:46] pieterh	however for the whitepapers it'll mean editing every single whitepaper to remove cruft
[18:46] pieterh	hmm...
[18:46] erickt	thanks pieterh
[18:47] erickt	i'm used to the docs enough that I'm not running into too many problems, I just figured any newbies would run into the same trouble
[18:48] pieterh	it'd be better to delete old stuff and move speculation to a separate category, don't you think?
[18:48] pieterh	i'll add the date/time to the category templates now
[18:48] pieterh	and then clean up the entire wiki later
[18:49] pieterh	there's no reason for 1.0 docs to still be there IMO
[18:51] erickt	sure that seems reasonable
[19:09] pieterh	sustrik: i've started a page to capture ideas for 0MQ/3.0
[19:10] pieterh	i assume API changes as you proposed for simpler message passing will require a major version
[19:10] sustrik	do we have a versioning policy?
[19:10] pieterh	http://www.zeromq.org/docs:3-0
[19:11] pieterh	i think there was consensus on versioning, yes
[19:11] sustrik	is it written down somewher
[19:11] pieterh	it's in the mailing list :-)
[19:11] sustrik	i, for example, have no idea what the outcome was
[19:11] sustrik	which email?
[19:11] pieterh	i was being somewhat facetious
[19:12] sustrik	i know
[19:12] pieterh	it is the list but that is not a database
[19:12] pieterh	i will write a versioning policy strawman, beat up as desired
[19:12] sustrik	ok
[19:12] pieterh	to cover the product version and API, not library versioning
[19:12] sustrik	yes, sure
[19:13] sustrik	library versioning is a different kind of beast
[19:13] sustrik	as for 3.0 i would like to have sockets to be represented as 'int' instead of 'void*'
[19:13] sustrik	not a pleasant change for users...
[19:15] sustrik	btw
[19:15] sustrik	http://www.zeromq.org/whitepapers:traffic-monitoring-v20
[19:15] sustrik	seems to have two tocs
[19:23] pieterh	all the whitepapers have two tocs until i get around to editing them all
[19:24] pieterh	i've started the policy page: http://www.zeromq.org/docs:policies
[19:24] pieterh	sockets as ints is a great idea and necessary for kernelization
[19:24] pieterh	would you add that to the 3.0 page?
[19:54] pieterh	sustrik: i'm cleaning up all the whitepapers
[20:04] sustrik	ok, will have a look
[20:05] sustrik	btw, mentioning author in the date header is confusing
[20:05] sustrik	erlag binding:
[20:05] sustrik	Written: 2010.05.15 by pieterh
[20:05] sustrik	Revised: 2010.08.4 by pieterh
[20:05] sustrik	date would do imo
[20:08] pieterh	ok, now i've gone through all the whitepapers
[20:08] pieterh	deleted (moved to deleted:whatever) the old binding descriptions
[20:08] pieterh	other historical ones are clearly marked
[20:08] pieterh	http://www.zeromq.org/navigator/order/title/category/whitepapers/
[20:08] pieterh	gives you a navigator
[20:08] pieterh	i'll remove the 'by' info from the templates
[20:09] pieterh	but the advantage is to know who to blame
[20:09] pieterh	please confirm you want that gone
[20:11] sustrik	well, you'll remove it yourself once people start spamming you about erlang binding
[20:12] pieterh	well, it's gone now
[20:13] pieterh	it is nicer to not have names overtly visible on every piece IMO
[20:13] pieterh	there is the history
[20:14] sustrik	my point was that creator & last editor != author
[20:14] sustrik	we may add author to the heading of each document
[20:14] pieterh	in most normal cases, last author and creator do represent the people to contact
[20:14] pieterh	this kind of mass update is very rare
[20:15] pieterh	it's only because we did not template the pages originally
[20:15] pieterh	templates allow a clean split between structure and content
[20:15] pieterh	the whole site is now properly templated i think
[20:15] sustrik	so why are you creator of erlang binding?
[20:15] pieterh	because i created that page :-)
[20:15] pieterh	obviously
[20:15] sustrik	you see
[20:16] sustrik	my point
[20:16] pieterh	no, normally it would show, accurately, person A creating, and person B maintaining
[20:16] pieterh	and you could ask person A, "why did you create the page HERE"
[20:16] pieterh	and person B, "why does it say XYZ?"
[20:16] sustrik	that's not the state of afairs now
[20:17] pieterh	indeed, because exceptionally, i had to restructure 50+ pages
[20:17] sustrik	it'll just mention either you or me
[20:17] sustrik	no much point
[20:17] pieterh	do you understand how the templates work?
[20:17] sustrik	yes, but look at the outcome
[20:17] pieterh	sigh
[20:17] sustrik	right now, those two fields are just garbage
[20:18] pieterh	this is a one off that will be fixed as soon as people start to edit their pages again
[20:18] pieterh	anyhow, it's moot
[20:18] pieterh	the names are gone
[20:18] sustrik	ok, let's add the two fields back once that happens
[20:18] pieterh	i'm only explaining their value in normal circumstances
[20:18] sustrik	sure, i understand
[20:19] pieterh	it's like saying "the knife cut me" and then arguing with the person telling you to not hold the sharp end
[20:19] pieterh	wikidot has powerful tools for doing this right
[20:19] pieterh	my fault for not doing this reorganization earlier
[20:20] sustrik	the point is that right now the normal circumstances are the norm
[20:20] sustrik	not the norm
[20:20] sustrik	so what you see is garbage
[20:20] sustrik	when normal circumstances become normal
[20:20] sustrik	it would be easy to modify the templete
[20:20] pieterh	indeed
[20:20] pieterh	templates are excellent
[20:20] pieterh	i strongly advise to never create wikidot pages without templates
[20:21] pieterh	its like trying to do everything using ZMQ_PAIR
[20:21] pieterh	mixed with semaphores
[20:22] pieterh	there is a long and somewhat painful discussion on HyBi
[20:22] pieterh	(IETF list discussing WebSockets)
[20:22] pieterh	including framing and multiplexing
[20:22] pieterh	i am trying to suggest 0MQ-style framing and delegating multiplexing to SCTP
[20:24] pieterh	i mention this because at some point 0MQ is going to bump gently against WebSockets
[20:27] Utopiah	the future of the intarwebs
[20:28] Utopiah	http://dev.w3.org/html5/websockets/
[20:28] Nwallins	they make webserver-initiated conversations (as opposed to client-initiated with HTTP req/resp) easier
[20:28] pieterh	it would be worth understanding websockets briefly, or at least the goals behind it
[20:28] pieterh	it is supposed to provide something as close as possible to TCP (async streaming) to browsers talking to servers
[20:29] sustrik	once i have some time free i'll have a look
[20:29] pieterh	its is pushed by WHATWG, representing browser builders, esp. Google
[20:29] pieterh	s/its/it's/
[20:29] pieterh	s/is//
[20:29] pieterh	i'll explain over our next beer
[20:29] sustrik	is there any one-to-many concept involved?
[20:30] pieterh	no
[20:30] pieterh	tcp
[20:30] pieterh	no patterns except peer
[20:30] sustrik	then it shouldn't be hard to layer 0mq semantics on top of it
[20:30] sustrik	the actual wire format doesn't matter that much
[20:30] pieterh	yes, that's for sure
[20:30] pieterh	but websockets risks being too close to the application
[20:41] sjampoo	Mixing websockets and zeromq is like a baileys cola. I don't think they go well together.
[20:42] sjampoo	Also, i'm not sure if there is much to be gained as it needs to be transported over HTTP anyway. The connection is always initiated from the client as a regular HTTP connection and can then be upgraded to allow streaming of data frames both ways.
[20:44] sjampoo	A ZeroMQ client implemented in JS would be cute, but you would end up needing a central (proxy) server anyway.
[20:46] sjampoo	Actually, If I had to pick I would say ZeroMQ is more like a Vodka, transparent, mixes well with everything and no bad hangovers.
[20:47] Nwallins	well, vodka cola sounds pretty bad ;)
[20:48] sjampoo	It is not that bad really ;)
[20:49] Nwallins	plus, martinis are high-powered, sophisticated tools that promote communication
[20:49] Nwallins	though I must admit a gin preference, there
[20:50] sjampoo	with tonic?
[20:50] Nwallins	just olives :)
[20:52] sjampoo	Hmmm, did you know that they say that Gin Tonic was the reason the british where able to invade Africa?
[20:52] Nwallins	the quinine, I heard
[20:52] sjampoo	Yes, it worked well against Malaria and the Gin made it drinkable.
[20:53] Nwallins	usually it's the other way around. mixer makes the drink palatable, and then you get sick ;)
[22:26] ateTate	I want to establish communication between two public webservers. What's a good way to ensure that messages recieved by my REP handler are from one of my servers and not someone else?
[22:28] Nwallins	http://lists.zeromq.org/pipermail/zeromq-dev/2010-July/004354.html