XMPP Summit - 2012-10-25

http://wiki.xmpp.org/web/Summit_12

http://wiki.xmpp.org/web/Summit_12 | remote participation: https://plus.google.com/hangouts/_/7a8d1536d98d4d57a9ebb42f89e771df1757d3d1?authuser=0&hl=en-US

hi Steffen

hi stpeter

is the conf. on?

we haven't quite started yet

waiting for people to come upstairs from breakfast

ahh ok. shouldn't it be now or ?

it should be now, yes

ahh ok.. here itøs dinner time.

it's

ok, enjoy your dinner

thats fine.. I'll take a quick bath and have a dinner. back in 20. is the conf. on then?

merci

stpeter, in retrospect i should have just used hangout on my mobile, i know its mic is working ^^

is hangout up and running?

Steffen Larsen, yup

see subject/topic or https://plus.google.com/hangouts/_/7a8d1536d98d4d57a9ebb42f89e771df1757d3d1

yes. i just followed the link

will join in 20 min

when will the summit start? or when is 9:00 in local time?

We're just having some network issues here, and we're missing a couple of people, but I figure we'll get started soon.

a link to the mentioned youtube broadcast, if it exist, would be nice too

We're using google hangouts right now, I believe.

yes

http://wiki.xmpp.org/web/Summit_12 | remote participation: https://plus.google.com/hangouts/_/b84e928d04633d26f5902ffcf6a597b4ba7fd3b3?authuser=0&hl=en-US

new hangout, got disconnected from the old: https://plus.google.com/hangouts/_/b84e928d04633d26f5902ffcf6a597b4ba7fd3b3?authuser=0&hl=en-US

changed it again: https://plus.google.com/hangouts/_/341b1d71f1b5294d0ef616e0dccee849aa918c1a?authuser=0&hl=en-US

http://wiki.xmpp.org/web/Summit_12 | remote participation: https://plus.google.com/hangouts/_/341b1d71f1b5294d0ef616e0dccee849aa918c1a?authuser=0&hl=en-US

I could not figure out the YouTube integration

ah..ok

There's always ustream if all else fails

stpeter, http://support.google.com/youtube/bin/static.py?hl=en&topic=2572818&guide=2572797&page=guide.cs you might already know that....

any remote folk speak up if they want me to relay questions or items

stpeter: ha ha.. nice summit on hangouts.. can you turn up or something? :-)

stpeter: the volume that is

it's the microphone on the mac - so the back-of-the-room folks may be low volume

we have to get a real mic. :-)

Heb, joining the hangout from my phone didn't work very well

*Heh

intro?

I couldn't hear earlier

brian beggs from EnerNOC

he gave a talk at the realtime conference - it will be up on the internet in a couple of days

(belated) grazé

y'all blinked! (-:

the hangout keeps dropping - peter is adjusting it when we hear the "drop" tone

Justin from fanout.io

could you tilt the screen back a little … he's headless (-:

how is that?

grazé!

I couldn't get his name, but he is from Erlang Solutions

/nod

michal slaski

!xep 286

xep-198

XEP-0238: XMPP Protocol Flows for Inter-Domain Federation

XEP-0286 is Mobile XMPP summary, basically. Power level background, and suggestions for implementations.

dammit we could use a real mic for the summit.. it's really hard to hear anything remotely.. :-)

we are turning the camera to make the mic better

bear: merci!

they are talking about transmit delay for bosh

We have a real Mike, if that helps.

please do speak up if you want to add

:)

yeah!.. can't at the moment, screaming baby, food etc. :-)

ahh

peter is adding to the whiteboard the items that need to be brought up at the bosh session

pirate pad http://piratepad.net/05gvwTiPKT

i'll be adding what is on the whiteboard to it

jack is setting up the projector for a presentation

https://plus.google.com/hangouts/_/341b1d71f1b5294d0ef616e0dccee849aa918c1a?authuser=0&hl=en-US

The link is in the room subject

Jack is starting his presentation

ralph is standing in front of jack with a tablet of the google hangout - jack is being very distracted by it :)

ha

http://www.html5gamedevelopment.org/StateofHTML5GameDevelopment

Ben Langfeld: I was just looking at that :)

http://metajack.im/tmp/crowdball/client.html

http://github.com/talkto/crowdball

http://display.ik.nu/xmpp

~/prosody/trunk$ wc -l plugins/mod_{message,presence}.lua 85 plugins/mod_message.lua 378 plugins/mod_presence.lua

lunch!

yes! lunch!

lunch? just had dinner ;-)

(-:

silly EU folks

I also humbly request y'all avoid summit topics in the lunchtime conversations (-:

lunch... mmmmm....

I will try to cancel any deep conversations

m&m, we'll just talk about you.

that works

dwd: I know you're jealous of my hair and beard

I am, it's true.

stpeter is really hungry… ha ha

thanks bear.. didn't notice!

hi peter...

hi!

seems to be a hickup

I think your video/audio froze :)

just did, yes

Steffen Larsen: actually I can go without lunch, but Dave Cridland was complaining :)

stpeter: ha ha, ok

stpeter: just look determined..

its not the skype codec..

Hi - joining late. Is there a video stream?

check the topic

google hangout

there is a hangout

topic

Florian linked me to this: http://www.youtube.com/watch?v=MyguMuZnHoc at 00:12

I try to rejoin the hangout, but I get the message that the hangout is over...

winfried: you have to click "yes" that you are still there in the hangout.

irritating.. but..

don't see the option at my google+ app for android... :-(

ohh prob. a small bug. ;-)

its there on a normal browser

I am back, but audio only

I have to apologize Stefan Strigler: I just got word he can't make it today

can somebody mirror the topic list?

http://piratepad.net/05gvwTiPKT

I am mirroring the pad peter is writing on

bear: cool

That URI asks me if I want to create the pad. :-/

I guess it won't work on Android's Chrome or something.

hmm

bear: double-you eff em

???

works for me (wfm)

Huh. Nor on the (old) Android Browser. Ah, well.

ahh

list as it stands:

1. BOSH vs WebSocket 2. XMPP over WebSocket 3. batch / buffer size 4. General BOSH cleanup 5. Session attachment 6. Shared BOSH / WebSocket

bear, It's OK, I can just read the flipchart from here, I just wanted to help type.

I keep forgetting some are in the room - brain is having trouble linking nicks here with real-life folk

Is this still the hangout URL? https://plus.google.com/hangouts/_/7a8d1536d98d4d57a9ebb42f89e771df1757d3d1 ? Says event ended.

pushp: https://plus.google.com/hangouts/_/341b1d71f1b5294d0ef616e0dccee849aa918c1a?authuser=0&hl=en-US

https://plus.google.com/hangouts/_/341b1d71f1b5294d0ef616e0dccee849aa918c1a?authuser=0&hl=en-US

Thanks waqas, stpeter

I missed about 45 seconds of that conversation

they are talking about the edge case that involves bosh receiving two requests with the same RID

and trying to find out what current implementations are doing to handle it

and we have come to a conclusion?

OpenFire must die?

Yes!

....no comment

Did you manage to get the feed onto something that doesn't require a google account to see?

getting very choppy sound

lost it here...

The solution we discussed for dupe RID was to return <body/> for the original request, discard data for the incoming duplicate request, and begin holding it.

thanks jack

(sorry for delay. had wifi problems for a bit)

One solution we discussed before was: auth with SASL, using the session cookie/token as password.

The web application becomes the SASL service which the XMPP server talks with, and the client need not go through the web app

choppy

beyond choppy.. disconnected

yea, the wifi is getting busy here

lost audio

MattJ bamf'd to a foreword seat

We have got two cameras going, could explain it - is anyone using Ralph's roving Nexus 7 cam?

reconnect - the audio is working to others

dropping and rejoining

everything dropped - reconnecting

wtf happened… well back again

now I am all alone in the hangout...

ha ha

our side dropped - we had to reconnect

I rejoined

I can see the others

maybe I am dropped again...

*sigh*

ouch

We have some implementations of that by the way. I can't recall, but it was probably Prosody's mod_auth_phpbb3, mod_auth_wordpress, etc.

of 'that' is: SASL external with the session id of the webapplication?

ask ralph to try landscape … it might provide a less squeezed view

We just used SASL PLAIN, with password=wordpress-auth-token. mod_auth_wordpress+Wordpress was an external SASL handler. No modification to the XMPP server or BOSH component.

better

¡gracias Ralph!

What Matthew Wild suggested also works across browsers (user has app open in multiple browsers/devices)

audio is super faint

and some extra feedback

I can't hear y'all

yikes

WTF

STAHP

auuuch

sound is f****

waqas: iirc you were originally on the shared-BOSH side :)

I missed the start of this discussion, was there discussed what would be the best level to solve this

?

MattJ: I was. I was also on the session-unbind support side.

unbind?

tofu: Multiple sessions on the same (BOSH) stream. You can ignore that :)

ralphm: don't quit your dayjob to be a camera operator (-:

I'm a little dizzy now (-:

m&m, Ralph says his arm got tired.

can't manage any more guys.. I'm off to bed. See you guys tomorrow!

nn

nighty nighty! :-)

'night

nn

http://caniuse.com/sharedworkers

Shared streaming documents. Client logs in, indicates what categories of documents it's interested in (MUC rooms, etc), servers sends all current state, and streams updates. :)

This is exactly what the roster protocol does

so... the idea is now a 'session proxy'...?

One of the ideas

brb

ralphm: https://www.erlang-solutions.com/products/mongooseim-massively-scalable-ejabberd-platform

brb - bb

Michal Slaski from Erlang Solutions

https://github.com/esl/ejabberd

jack: ^

are there slides?

(-:

:)

of course not

why would you want that? ;)

Which badly hurts BOSH performance

if you want long connections, why don't you use normal TCP? :P

Tobias: Because the W3C+IETF wouldn't let us

yeah...that bad IETF :) aron didn't bash about w3c a bit....didn't they have their fingers in oauth too?

You can't blame them. Raw TCP isn't WebScale.

He said most stuff went wrong under the IETF

waqas, only node is webscale

doesn't w3c have even more coporate participation than ietf?

There's Tigase, with their pseudo-XML parser :)

;-)

btw sorry for forgetting to mute myself after relogging in

I need to drop out … could you post the followup discussions on the list before you all adjourn?

sure can

great!

I remember a blog by Jack on translating it...

http://metajack.im/2010/02/01/json-versus-xml-not-as-simple-as-you-think/

JSON-LD supposedly solves the namespacing issue

but the array vs. non-array thing is not as easy

yep

or perhaps i don't know enough about json-ld

i should probably write a version 2 of that blog comparing it against json-ld

http://json-ld.org/spec/latest/

client side serializing/deserializing? makes me shiver...

+1

Every problem can be solved with another layer of indirection

but what is the problem? webdevelopers not able to use a api?

Apparently

web developers are like school children and XML is like cooties

hahaha

Does this surprise anyone?

i'l avoid saying things about php and javascript-in-browser...

bear: tempting wasn't it

*very*

{ "iq": { "-type": "set", "-from": "hamlet@denmark.lit/blogbot", "-to": "pubsub.shakespeare.lit", "-id": "pub1", "pubsub": { "-xmlns": "http://jabber.org/protocol/pubsub", "publish": { "-node": "princely_musings", "item": { "entry": { "-xmlns": "http://www.w3.org/2005/Atom", "title": "Soliloquy", "summary": " To be, or not to be: that is the question: Whether 'tis nobler in the mind to suffer The slings and arrows of outrageous fortune, Or to take arms against a sea of troubles, And by opposing end them? ", "link": { "-rel": "alternate", "-type": "text/html", "-href": "http://denmark.lit/2003/12/13/atom03" }, "id": "tag:denmark.lit,2003:entry-32397", "published": "2003-12-13T18:30:02Z", "updated": "2003-12-13T18:30:02Z" } } } } } }

(http://jsontoxml.utilities-online.info/)

that's suprisingly straightforward

why do some start with -?

attributes

(brb)

attributes to the element

ralphm: now do one using xml:lang

(ba)

of course there are some issues. Some elements *may* actually be specified more than once at a particular level. E.g. in this example, there might be multiple "link" entries, and that translator the wraps those in a list: link: [{"-rel":....},{"-rel":....}]

Lance: I assume it will just do '-xml:lang'

ralphm I was thinking of the matter that you can have multiple of the same item as long as the lang is different

one issue is that youcan't do link.append("foo") because it may be a list or it may be a single element

so they all have to be lists right?

jack: in that case it does become ugly

jack: and that's when bear's suggestion becomes more attractive to me

what was bear's suggestion?

fixed model, loses extensibility

in the general sense

seems like a nonstarter though. won't you devolve into IRC like stuff where people text encode the customizations inside the "standard" elements provided?

{http://jabber.org/protocol/disco#info}query -> "disco_info"

Or a fixed model which decorates an XML model, gaining both benefits

a mapping like that

Fixed model for known items, stock tools for JSON-a-like payloads, and we didn't discuss extensions to the fixed models.

I take an XML node and I decorate it with a dev-friendly model.

I can still do all of the generic XML operations I enjoy, and everyone whose use-cases fit the decorator can just use that.

Ben Langfeld: can you paste a small example?

roster.add("Dave Cridland", "dwd@dave.cridland.net").x("group").t("Fools")

Presumably.

presence = Blather::Stanza::Presence.new presence.status = :dnd presence.add_child 'foo' # => <presence><show>dnd</show><foo/></presence>

Dave's example is also the same idea

Ben Langfeld: +1 that's what sleek does, especially since it followed blather's model

dave's example would fit the nodeJS folks to a T

<off-topic>Blather does this badly right now but I will fix it</off-topic>

bear: almost, they'd want: roster.add({name: 'Dave Cridland', jid: 'dwd@dave.cridland.net', groups: ['Fools']})

roster is a bad example, as we don't do extensions there

lance: wouldn't the .().().() pattern fit better?

nope, you put all the optional args in a dict. that's the established pattern

Yep

Blather and Sleek do this

ralpm, Yeah - when didwe decide not to do extensions there?

My example was adding the group element manually, of course.

dwd: not sure, but not regretting it :-)

Or pretending to - I don't think it could work like that.

lance: oh, that didn't look like what substack was showing - guess I still don't get NodeJS

ralphm, Oh, I quite liked the idea. I implemented it before finding I couldn't find the spec for it anywhere for the commit log.

bear: what he did was for stuff that actually needed to pipeline. you don't need to pipeline to create a roster item

If anyone likes Objective-* examples: https://github.com/ArchipelProject/StropheCappuccino/blob/master/TNStropheRoster.j

ralphm, In fact, M-Link tags autogroups (shared groups) using extension elements inside roster items, actually. Works fine.

dwd: oh, ok

I would say that the big majority of XEPs actually have pretty fixed schemas that don't allow any extensibiity

Seconded

ralphm, But I *thought* we had arbitrary elements (and a requirement to persist) in RFC 6120 somewhere. It's nowhere there now, so I guess we removed it entirely at some point.

exceptions are message stanzas and pubsub payloads

and error conditions

occasionally we add new xeps that retro-extennd xeps. like rsm, shim, etc

ralphm, Sure, but this turns out to be really quite useful. I suppose it needs a disco feature and things. Maybe I should write a XEP.

the main thing is that we do namespace versioning, but that's not too hard to deal with

Lance: agreed, but you could possibly wrap those and it is usually orthogonal

To the general point, I don't see that the wire format is the problem so much as libraries being too general and assuming that end-users will read XEPs and construct stanzas manually.

Which…they won't

ever

socket.io and the like are successful because the API is like 1 method

+1 Ben Langfeld

+1

I've seen people complain about libraries not implementing usertune, so yes :)

indeed

Ben Langfeld: agreed. I just wanted to put it out there. If you don't need generic mapping and/or stanza construction, then you basically need to define a model that maps the existing xml serialization, whether or not you define an actual json serialization.

Why is no one mentioning AMQP, which is the industry standard with multiple implementations, as opposed to 0mq which is basic? :)

we have

I missed that

AMQP has been mentioned a couple of times

the folks who wrote 0mq were around when amqp was written

The one thing amqp has going for it is reliability guarantees as part of the protocol. I don't think any competing protocols have that.

before everybody runs of to the dinner... maybe good to wrap a few things up on BOSH

open ends from todays discussion (imho);

RIDs: who writes the patch?

HTTP/XMPP authentication: is there enough consensus to write a protocol/extention for it?

winfried, Yeah, I think there was enough consensus to work on that.

then of course the next question: who will pick it up? :-D

Yeah, not sure we have an actual volunteer for that. And Lance is back, so we can't volunteer him again. :-)

wait

again?

then on websockets: is there any need for work on the proposal or are there extra things to standardize, or does the draft from Jack and Eric Cestari work for now?

(I think it does work right now, but better check it)

BTW, the brownies and cookies are nice.

<3 cookies

plz safe some of them for me ;-)

>> cookie monster

waqas: zeromq is misnamed. it has no relation to amqp except hte people behind it and the "queue" in th ename

it violates the liskov substitution principle in both directions :)

while considering bosh & websockets, should we do a binding for server sent events? sse feed + post endpoint for sending sounds like it should work rather simply.

jack: I know. I initially missed references to amqp and thought only 0mq was being discussed.

we were mixing it all together :)

still are :)

m&m: is that still stpeter's office?

redecorated

m&m, Move everything around on the shelves.

nope

that's my home … you're seeing a small subset of my LEGO Star Wars collection