XMPP Summit - 2011-02-07

morning all

morning

morning

Xavier said via email that he enjoyed giving his talk and appreciated the people in the room who were helping organize

bear: good -- please ask him to send us his slides :)

Good morning everyone.

Can someone post the link to Peter's i18n presentation from Friday.

☕

Wayne Franklin: http://www.saint-andre.com/jabber/i18n-for-the-perplexed.pdf

good morning

Aloha

Morning

would love to see a recording of that i18n talk - I can imagine more talking happened than what the slides show for words

bear: I don't think we recorded the first talk … which may leave you behind with the second /-:

thats fine - i'm sadly a very broken and scarred unicode veteran ;)

so the second talk may be more useful :)

Some other characters to add to your list of edge cases: ﷺ (U+FDFA), ΐ (U+0390), Dž (U+01C5)

hildjj: cool, let's keep track of those -- perhaps we want to create a wiki page on ietf.org?

nod.

reminder: http://www.mclean.net.nz/ucf/ for those without unicode checker

We control the temp by changing the number of hackers in the room.

stpeter: do we already take care of ZWNBSP?

http://typewith.me/AzdW3AgDSZ

Urgh.

So, how do I usefully contribute to this from here?

Hey Kev

How are you feeling?

I'm out of bed, which is an improvement on most of the last week, thanks.

I'm not so bad now.

You can follow the transcript thingy stpeter is keeping at the URL above

and type here

Yep, I'm doing that.

There seem to be three main requirements for DMUC - continuing access during partitioning, dealing with slow links and reducing ~duplicate stanzas across a single link.

Dave is talking about DMUC.

i would love it if he burst out in song, of course.

I'll wait for someone to transcribe, then.

Oh, you're in Cisco. I don't suppose the room's equipped to webex is it? :)

Kev, we looked into it, but couldn't make it happen here.

Bugger.

it's not our home office, so we couldn't make destructive changes.

Kev: Don't say that near Fritzy

Kev: Dave can bring you up to speed :)

MattJ: fritzy hates teh webexes?

ClusterMUC™

hildjj: Nah, Kev's exclamation of defeat

could folks in the physical room help scribe more either here or in the TypeWith.me page?

hildjj: +1

could someone repost the URL?

http://typewith.me/AzdW3AgDSZ

http://typewith.me/AzdW3AgDSZ

I'd greatly appreciate it, I'm not happy about have to have missed this.

Kev: you're already punished enough because you had to miss the pizza

talking of food. are there any plans to catch some food during the summit?

js: There shall be lunch

I believe there will be food.

lunch will be served at noon

nice

Several to Many.

ralphm@ik.nu: The slides are in stpeter's email yesterday - 'my slides for Monday morning'

MattJ: thanks

+1

It's not mine, as such. it's mostly Curtis and Kev.

Some things need hashing out though

It's also much better than IRC at netsplits :)

Kev: do you want a live video stream?

Do you mean something more advanced than a Skype call?

ah, you have that already...

I would offer Qik or uStream

florian is sitting closer to the front than i am, so his might be better anyway.

Indeed, why setup a complicated Proxy or master-slave system just to mimic IRC-netsplits ;-)

MattJ, I certainly agree there's scant detail in what we're proposing, as yet.

I'm hearing Peter fine atm.

Dave just had a netsplit I believe ;-)

give me 10mins to set it up

So my current state here is that I've got a new MUC-Sync spec I need to write, and then update FMUC to use it, and then discuss it.

winfried, Yeah... Need a local server.

So I'm happy to do it that way.

Because no-one's insane enough to want to define MUC 3.0 based on pubsub.

And the issue with sync etc. is exactly the same.

someone is streaming?

I will be in a second

I'm on a Skype call with hildjj.

just figuring out Qik on the N900

ah

My attention span is no longer keeping up with my fingers. I'm dropping off typing on typewith for a while

Thanks Ben, it's been useful.

does the older flood/fill algorithm of olden internet days useful here?

wow - early morning english ftw

hildjj: could you get me a second WiFi login for the N900 to stream?

wifi must be dropping folks a lot in that room

works fine with 5 GHz

bear: Really? :)

I guess Joe's trying to stream to Bear as well as me now, and it's killing his connection.

kev/bear, i'm going to just start a webex.

It was fine previously.

yea, i'm going to drop off if it's causing issues with kev's reception

though on friday it had a few problems with 2.4 GHz

Streaming video would seem interesting, if we can.

i'm on a different subnet/radio channel than you guys.

I get really bad wifi

https://go.webex.com/go/e.php?AT=WMI&EventID=138622147&PW=660e3d5d0402060b530e&RT=MiM2

florian: try 5 GHz

hildjj: Thanks very much, going through the Preparing dance now.

http://qik.ly/GfPtx

bear: put FF4 into 32bit mode.

Qik should be streaming

It is, thanks.

I'm streaming both atm so I can swap when one fails :)

low quality as the WiFi doesn't work for me

is the quality ok?

Florian: I can set you up for wifi

I have WiFi keys, they network in this corner just doesn't work properly

will try to improve it for the second half of the day

ah

coffe cup moved :)

is it just me that is having issues with the WiFi dropping packets massively?

well, I guess that Kick just answered my question :)

http://en.wikipedia.org/wiki/Operation_Market_Garden

I've got webex and qik going, one on each of two machines, and I'm turning audio on/off on each stream based on who's speaking so I can hear them :)

hehe :)

http://twitter.com/#!/mobilevikings/status/34559653098561536

;)

Fun.

btw is it possible to get another wifi key for my mobile?

my problem was my server's route being lost, not the wifi :/

time for a break

Can we have a live stream of the sandwiches? :D

Kev ... sure

i'll set one up :D

Not really.

@Kev: we can try to stream them across the channel...

heh

for this afternoon's session, we'll be chatting in xmpp:xmpp@jabber.ietf.org

hi bear

hello again

black ice and drunk driver == power pole destroyed

ouch

and me finding out I have one UPS overloaded (the hard way)

https://go.webex.com/go/e.php?AT=WMI&EventID=138622947&PW=252d4c05551816160222&RT=MiMyMg%3D%3D

and please move to xmpp:xmpp@jabber.ietf.org?join

*wave

hi

hi

we're all over in xmpp@jabber.ietf.org

i know

I have got some s2s issue it seams, can't join the room...

same

investigating...

what soft is running there?

Artur Hefczyc: we're running Tigase :)

LOL

hehe

good

what version of the Tigase then?

Artur Hefczyc: see my private messages :)

Name: Tigase Version: 5.1.0-b2506 Os: Linux-amd64-2.6.34.6-xxxx-std-ipv6-64, Java HotSpot(TM) 64-Bit Server VM-17.1-b03-Sun Microsystems Inc.

I asked what software is running on jabber.ietf.org

!version jabber.ietf.org

Dave Cridland: jabber.ietf.org doesn't support feature requests

!version ietf.org

Dave Cridland: ietf.org is running ejabberd version 2.0.1 on unix/linux 2.6.22

Kanchil, Thanks.

ah, ok

So not my fault. :-)

hehe

It's nice for a change.

that the beauty of putting your soft on jabber.org, everybody blames you ;-)

!version summit@muc.xmpp.org/Astro

Astro: You are running Gajim version 0.14.0.1-c63b9caadd5b on Debian GNU/Linux 6.0 (squeeze)

hrm, did we implement version in the end?

!version robert.mcqueen@collabora.co.uk

robert.mcqueen: robert.mcqueen@collabora.co.uk doesn't reply to version requests

robert.mcqueen: full jid?

!version summit@muc.xmpp.org/robert.mcqueen

robert.mcqueen: summit@muc.xmpp.org/robert.mcqueen doesn't reply to version requests

bah

.

no, we didn't. we talked about a possible implementation on Friday, remember? :)

any advances Florian? getting a kind of impatient over here ;-)

give me a sec please

looks like ejabberd is broken :)

It's working okay for me.

apparently you have broken soft too ;-)

have seen interoperability issues between tigase en ejabberd before

(our Berlin-branch runs eJabberd ;-) _

winfried: sorry, I am working on this now, but it looks it may take a while before I can fix it

or workaround it, whatever it is

ok...

thanks for diving into it Arthur!

it's like never ending story between tigase and ejabberd, once I get it working, they release a new version which does not work :-(

Artur you are being picky :)

yes, you are right

sorry

I know, that the Tigase which is buggy and needs fixing, I should get back to work on it and stop complaining

:-)

well... complaining seems to be fashionable

never considered pointing out failure patterns to be complaining - but that's just me I think

indeed

eh, and I was so close to the ultimate solution for all world problems and Tigase performance, but…. I was called to presentation and now it is all gone ;-)

42 it is ;-)

lol :D

low battery, leaving

hildjj: Why does the I and V not recompose to "IV"?

MattJ, because there's no recombination specified for I + V, or Rajiv would recompose.

which would be bad.

because you'r loosing information. I V -> I V, IV -> I V, now what should be the recompose of I V ? S.th. from {I V, IV} of course. And the simplier one wins....

hildjj: Oh, so they don't decompose to something like roman numeral I, roman numeral V

right. they decompose into a "normal" i and v.

https://secure.wikimedia.org/wikipedia/en/wiki/Pigeonhole_principle

arty: Yes :)

I was just surprised that such a simple character could grow in "width" after NFKC

there are some little nuggets still coming out.

Oh, this isn't the room I thought it was

Do we have a remote clue about what would be needed for international usage or is this about enabling and blocking random char sets?

I kinda agree: we need some more international people here before we can tease this out more. This stuff is complicated.

arty: There are sets of unicode characters marked as e.g. "symbols" that we could choose to drop

I'm not in favour of that though

yes, or we could choose to support what people *care* about.

we also have the issue of whether this stuff is even displayable on clients. For example the "pile of poo" is not even displayable - http://www.fileformat.info/info/unicode/char/1f4a9/browsertest.htm

The thing I remember about bi-di is that you have to deal with inserting western language in the middle and the western stuff has to read left to right even though the flow of a sentence is right to left.

about the numbers: arabic numbers for example are just LTR characters...

SimonT: wtf :)

For a short time in North Carolina they were issuing license plates in the series WTF-####

arty: the problem is that "what people care about" can change, and what we decide now won't be changed for decades ;)

Wayne Franklin: lets say for example lowercase is LTR and uppercase is RTL then the string is stored like: abcDEFghi, but rendered liek abcFEDghi

https://tools.ietf.org/html/rfc5892

so that is a UI issue, (when embedding those in html, it can mess up your rendering), but can't think of an issues with XMPP

example from RFC 3454, section 6: If a string contains any RandALCat character, the string MUST NOT contain any LCat character

If a string contains any RandALCat character, a RandALCat character MUST be the first character of the string, and a RandALCat character MUST be the last character of the string

That's one way to deal with it

MattJ: Well, societies usually move slowly. Anyway, I'm not sure if I want to see Šƫǝƒǡɳ in my contacts list, whereas I accept Константин Тарасов ..... I'm just not sure if the discussion at this abstract technical level is in any way correct or respecting reality

and there's also U+200E and U+200F.

arty: The problem is that (unlike IDNA's registrar-blocking approach) we can't really control things geographically the same way

I don't want to accept Константин Тарасов - I'm a stupid Englishman :)

:)

:-)

This is all very localised, but XMPP is global and decentralised, we need to find a common ground on what we can/can't globally enforce

I have our admin working on making sure everyone in the world just speaks English.

actually, you don't have to accept him, but you should allow him to use his native language for his name.

I don't think it's a bad idea to follow the example of email addresses for JIDs

hildjj: he's root, he can do that...

arty: Many of the characters in "Šƫǝƒǡɳ" are "native" to some part of the world, is my point

well she's relatively non-technical, but very persuasive.

Blocking stuff like this isn't really a protocol issue

hildjj: U+200E and U+200F are a problem indeed

What are U+200E and U+200F? Space characters?

Wayne Franklin: LTR/RTL

oh

LEFT-TO-RIGHT MARK, RIGHT-TO-LEFT MARK

MattJ: do you know or asume that? It would be really easy to e.g. pull all wikipedias and check all names in all languages (quite easy, did more or less that for en.wp.org)

arty: caron is commonly used in languages, sure

I think the others are mostly phonetic

But my point is that we shouldn't randomly block things that look "strange" to us

because later on someone will come along and find that due to our ignorance in 2011 they can't register a user account on a server in their native language in 2020

whateve

r

That's why I mean measure it. Analysis of wikipedia is a bit biased but fast & easy, and perhaps safer than doing for some random unicode flags

and it might help us to understand the problem space better

arty: suggest that on the list.

Measuring wikipedia isn't exactly scientific

MattJ: Suggest that on wikipedia-research list :-)

I agree it's better than making uninformed decisions about what we block, but then - why should we block anything?

Many of the "valid" characters are as, or more, complex cases than symbols and what-not

Except for really obvious cases like direction markers and some space stuff, I'm not sure there's a point in trying to decide what should/should not be allowed in XMPP

My gut feeling is that we need a normalization form (and mapping rules for case folding), but I suspect aside from things like space characters not being allowed, much will end up as local policy.

and obviously even direction markers are up for questioning

Dave Cridland: That's not easy if you then get interop issues between servers

arty: Measuring wikipedia gives you statistics, statistics are not science, they're a sub-class of lies as everyone knows

MattJ, Assuming that servers normalize on the wire, and domains are normalized and mapped interoperably, I don't think there's a problem.

Normalization is what we are discussing, no?

MattJ, Only partly. Normalization is basically NF[K]{C|D} + case folding.

Sure

so if you allow one set of characters, and I don't... I'm meant to bend my normalization rules?

MattJ, Servers *may* also need to consider the RTL/LTR issues, and the rest. It's not clear to me that a global policy on anything but domain parts is required.

I don't see that things are going to work if we have different normalization rules on different servers

Disallowing characters isn't part of normalization, is it?

Kev: Technically I'd say not, but for the purposes of this discussion... :)

it's covered by NF*

aiui

MattJ: So what would you measure? Wikipedia has especially person names. Many person names. How about just extracting them and see how they work against suggestions? Yes, this will not be the final result but a nice *testset*. I dislike settling without any sort of idea what would be needed.

I mean, given that I use NFC (for example) for normalization, doomsong my still choose to disallow local localparts that have a 'z' in them.

MattJ, NF* covers prohibited characters? I didn't think so.

That doesn't cause interop problems with jabber.org, which naively allows people with z in their name.

arty: I dislike setting, so we're kind of in agreement :)

MattJ: Are you in favor of case folding?

Yes

Kev, Right, and in addition doomsong.co.uk would only care about it's own localparts.

Wayne Franklin, We have case-folding on local parts already, i think changing would be catastrophic.

Kev: You can decide on your server to allow/disallow what you want - but we're setting a baseline that everyone MUST comply with

Well, it cares in three ways. It cares about applying normalization everywhere, global prohibitions everywhere, and local prohibitions (z) only to local addresses.

I'm just trying to find where the line is to be drawn between accepting and not accepting a jid part

saying "there is no baseline, do what you want" is a recipe for disaster as far as JID internationalisation is concerned

MattJ: Yes, and that's just fine, but the question is whether the baseline has to be very high at all.

You have to prohibit anything resembling a space, certainly.

Right, that's what I think too

I'm not in favour of blocking large ranges of characters

MattJ, So what the idea is is that locally, you might say "My users may not have half-width hangul characters".

I am, but I'm in favour of the servers doing it, I think, rather than protocol.

I don't see why a server would /want/ to, but they're free to do that as a deployment decision if they wish and accept the consequences

aren't we already screwed - does XMPP disallow xn-- domain parts atm?

so people can and will set up jids @ localised domains

which will go onto rosters...

MattJ, But globally, we can say "Take a node, apply NFD, case-fold, and if it contains any space-like characters reject"

so we basically already must require that, for example, servers must decompose domain parts to punycode for storage/comparison

Robot101: We don't use the xn-- domains directly, their unicode equivalent go across the wire and in storage

or we'll never have any backwards compatibility

Robot101: conversion to ascii is only for DNS lookup

what if someone configures their server to be @xn--.lala ?

Robot101: You get what you deserve ;)

Robot101: If you do that in prosody then it will return "host-unknown" when someone tries to connect to your (unicode) host

is that documented anywhere as "don't do that" - in terms of hildjj's concern of people complaining at us?

maybe its ok because its not a new problem

I don't think it's documented, it's just not your hostname so you don't configure your server that way

your hostname is the unicode variant

hildjj: minutes are here: http://typewith.me/oVoICs8hD5

could everyone at the end of this event give me back the SIM cards, as I need to send them back