Zashfritzy: Fine, thanks. I arrive. Except now we're out of coffee. I may have had something to do with that...
Steffen Larsenhas left
bearcan you hear him?
ZashKev: Tobias did implement -PLUS, but it has not been merged yet
Steffen Larsenhas joined
Adrian Georgescuhas joined
bearlance - what is your jid? I just realized I don't have you in my roster
Steffen LarsenWebRTC and MUC: http://candy-chat.github.com/candy-webrtc/
Dave CridlandAlexey tried to make DIGEST-MD5 use saslprep, but as I recall is was a rathole.
hildjjdave: that's my recollection. the whole idea of having a separate *prep for SASL vs. your application protocol is a little wonky anyway.
Dave Cridlandhildjj, Well, given that SASL authentication identifiers are already technically unrelated to the authorization identifiers from the protocol, maybe it makes some weird kind of sense.
Dave CridlandNot that I can really hear much of what's said, but Kurt Zeilenga's take here is that it makes sense to have multiple jids, one in ASCII and one not, and either have some mechanism to find out the real non-ASCII jid from the ASCII handle, or else have the server actually remember which jid alias to use for which contact.
ZashDid you just say 'jid alias'?
Dave CridlandZash, Yes, but actually in the sense I used it I mean jid equivalent.
KevDave Cridland: Yes, but that's to solve a different problem.
Dave CridlandKev, Ah, OK.
KevKurt's argument for having an ascii and a non-ascii is so you can put easy things on your business card or whatever.
KevRather than suggesting that Japanese folks need to have an ASCII JID.
Dave CridlandKev, Right.
Dave CridlandKev, But ISTR his general goal is remote canonicalization.
ralphmwith something like punycode?
Dave CridlandWhere is the microphone? Because I now can't here Kev, even. Is it shielded by the laptop lids?
KevI'm just not shouting any more. I was making an effort TO PROJECT earlier.
Dave CridlandKev, You is becomin' an hack-toor?
ralphmCan you hear Kev now?
Dave CridlandVery much so.
Dave CridlandIf he projects much more we can drop the webex, I think.
ralphmhe's presenting now
Dave CridlandOf course, I can't see his diagram.
Dave CridlandYes. Though for a brief moment I could only see his crotch.
ZashWhat darkrain said
darkrainhere's a muc, there's a muc. everyone muc muc
Lloydyou beat me to that joke :)
darkrainGreat minds and all that.
Dave CridlandThese two mucs count as one muc.
ZashDid Kev say MUCeption?
ZashEverything needs more MAM!
Dave CridlandYou can turn me round again now.
fritzyEnd <- encryption -> End
Dave CridlandWe have a discussion about e2e encryption every time, so should have one this time if only for tradition's sake.
Lancecan someone just tell us how to do it properly this time?
SimonTennantSomewhat related to Federated MUC is the Zookeeper project ZAB: http://research.yahoo.com/files/ladis08.pdf
Dave CridlandLance, The encryption bit is easy, it's the authentication and key exchange bit that's so damn hard.
fritzydid we lose audio?
fritzyoh, there it is
Dave Cridlandfritzy, They lost speech.
fritzyI have a lot of opinions on xmpp as middleware
fritzyok, I just have one big one.
fritzyDon't do it unless you're federating data.
Dave CridlandI think it also depends on where your middle is.
Dave CridlandXMPP does to the auth stuff pretty well.
Dave Cridlanddoes do.
fritzyright, but middleware is often network secured rather than identity authed
Steffen Larsenfritzy: no thats not always true..
fritzyoften, as I said
Dave CridlandRight, purely internal middleware is not interesting, but if your middleware is talking over the internet, say, that changes things whether federation is in play or not.
fritzydon't bother unless it's going over the public net
Dave CridlandI personally suspect that middleware is talking over mobile links and direct-to-browser a lot more than it used to.
fritzysure sure... clients are further out
Lancedave: yes, it is
fritzysocial network: might wait for Simon tomorrow?
Kevfritzy: He's here.
Dave Cridlandfritzy, I think he's that collection of eight pixels towards the end on the left.
fritzythis cisco tech is really amazing
Dave CridlandIt's like being in the room.
Dave CridlandIn a parallel universe where you see in 8-bit games console graphics.
fritzyyeah well, most of it is filled in by my imagination and remembering actually being in that room
Dave CridlandFlorian, You can't laugh, you weren't born when we had graphics this bad.
fritzyit's like that dream where you can't see anyone's faces
fritzythis staying awake all night thing was easier when I was younger
fritzyI think if we solve the web-integration problems, that the social network stuff will come
fritzybecause the social network people will come
Dave Cridlandfritzy, I think that's certainly a lot of it, yes.
hildjji'd be happy to shut the camera off if it's not helpful.
fritzyhildjj: what would we complain about then?! ;)
hildjjthere's no camera that could shoot this room.
fritzyit's helpful. Thanks Joe. Sorry for complaining
Lanceright. the problem right now is that it is too simple and easy to build a silo system, compared to integrating xmpp
hildjjthe lighting also sucks, which doesn't help.
hildjjit's like WebRTC, but with more angle brackets, and a bunch of namespace stuff you won't understand.
Edwin Monshas joined
KevBut it's OK, we can show you how to copy/paste.
fritzyright, we have a strong sense of identity
fritzyI repeat... if we solve the web integration problem, they'll come.
darkrainThere's an inherent "that XML stuff is scary; I've seen SOAP" attitude, perhaps
Dave Cridlandfritzy, Universal identity and addressability, no less.
hildjjdarkrain: yes. but i'm tired of fighting that battle.
hildjjpartially because i agree with them. XML blows.
fritzyit's a battle not worth fighting anymore.
fritzytime for an attitude change
darkrainI agree; just need to find a way to side-step it.
hildjjright. i've been at least as much of a problem as anyone else for maintaining the status quo, and i'm trying to not be anymore.
SimonTennantCridland: "XMPP: Universal identity and addressability" sounds great.
hildjjsimon, i usually say "identity-based routing"
Steffen Larsenid, presence and routing.. thats how I use it as middleware..
sekistnerWhat about scaling, how do you tackle bottlenecks in a federated system?
hildjjsekistner: it's always the other guys problem. :)
Dave CridlandSimonTennant, Yeah, it's a bit sound-bite-y, but it's the basic thing - every entity in an XMPP network, from the user (or their account in abstract), and each connected client or server, are all given an addressable name.
fritzysekistner: implementation problem... and implementations have solved it.
Dave Cridlandsekistner, I'm not sure I understand - when a system is highly federated, it tends to scale better (because other people take on more burden).
Adrian Georgescuhas left
fritzywebrtc people.... GET OUT
Steffen Larsenget a room
hildjjok, i'm going to drop from webex. peter can crank the a/v back up.
sekistnerNot in the web usually. If twitter was federated a retweet from Justin bieber kills a small server.
Dave CridlandOh. Suddenly, I can see PSA's desktop.
Dave CridlandYeah, I heard Bacon...
stpeterwhat do you think aboutmoving item retrieval to the extensions spec?
Dave CridlandBut I wasn't really paying attention.
Dave CridlandOK, I'm all in favour of moving things out of XEP-0060.
fritzyso we're slimming pubsub out?
Dave CridlandI'm not in favour of moving them into another kitchen sink spec.
fritzywe could do a "non-recording" version of pubsub
fritzyand then have a storage/retrieval extension
ZashIs there a persistence feature?
fritzynormally, yes... the idea might be to move that to its own extension
Dave CridlandI've never been sure what "persistence" means in terms of pubsub, actually.
fritzyitems being retrievable
fritzybeyond the initial notify message
Dave Cridlandfritzy, No, I don't think so. I think that's controlled by max_items, not by persistence per se.
Dave CridlandSO not about node lifetime at all?
fritzysure.. but they're talking about max_items, and persistence and retrieval being on the side
Dave CridlandRIght, I could go along with that.
ZashRight, http://jabber.org/protocol/pubsub#persistent-items was the disco feature
Dave CridlandFWIW, one thing I've seen from being engaged in sales calls about pubsub is that every single customer has a different set of requirements for a pubsub service, and there's too many possible variations.
hildjjEveryone not in the A/V breakout, please ignore this:
Dave Cridlandhildjj, Tunnelling SIP inside XMPP verbatim?
hildjjSIP over XMPP: SoX
fippodave: you're not supposed to listen to this ;-)
Dave Cridlandfippo, No, because it makes my brain hurt. I can (readily) go along with SDP tunelling; but throwing out Jingle entirely in favour of SIP doesn't strike me as a step forward.
stpeterDave Cridland: SDP tunneling makes some sense for various reasons, including CUSAX clients (which we'll discuss tomorrow when Emil is here)
Adrian GeorgescuAddressing for federated service
This specification describes how clients can discover multiple
services to configure themselves with a minimum of user-provided
information, as short as possible sequence of queries and with a
minimum of overhead for administrators of the services.