[18:33:43] --- arifumi@2entwine.net has joined
[18:33:55] --- arifumi@2entwine.net has left
[18:41:12] --- lars has joined
[18:42:28] --- lars has left
[18:43:51] --- lars has joined
[18:44:42] --- arifumi@2entwine.net has joined
[18:45:36] --- brabson has joined
[18:45:52] --- csp has joined
[18:46:02] --- brabson has left
[18:50:01] --- brabson has joined
[18:50:09] --- sarolaht has joined
[18:51:39] --- wouter has joined
[18:56:04] --- lahey has joined
[18:58:17] --- wouter has left: Logged out
[19:02:48] --- brabson has left
[19:11:13] --- magnus has joined
[19:16:08] --- magnus has left
[19:19:49] --- magnus has joined
[19:22:19] --- magnus has left
[19:26:06] --- magnus has joined
[19:26:42] --- kohler has joined
[19:27:43] --- magnus has left
[19:30:46] <kohler> Seems fine to have general service codes, other apps can define their own, RTP* codes would be ok as backup
[19:31:31] --- magnus has joined
[19:34:42] --- magnus has left
[19:34:51] --- csp has left
[19:37:04] --- csp has joined
[19:37:57] <csp> ok - thanks Eddie.
[19:38:14] <kohler> no problem! looks good.
[19:38:25] <kohler> i'll prob have more comments on it later
[19:38:37] <csp> ack. look forward to them
[19:41:24] --- lahey has left
[19:42:04] <kohler> is sctp multihoming and mobility now considered a mistake? dccp's support in this draft is basically analogous
[19:42:17] <csp> will ask
[19:42:25] <kohler> thanks!
[19:44:31] <kohler> this is nat safe -- it's the first dccp version that is
[19:45:34] --- magnus has joined
[19:45:56] <kohler> can't hear al
[19:46:11] <csp> better?
[19:46:14] <kohler> yes
[19:46:19] <kohler> thx
[19:46:47] <lars> michael tuexen
[19:46:52] <kohler> y
[19:47:38] <kohler> happy to change the name of the draft to "multiple addresses"
[19:50:36] <kohler> this is backward compatible, all done through options; maybe when pasi is done
[19:53:23] <lars> eddie: the draft says this can' t be easily done at the app layer. but isn't this equivalent to Mobile Sockets, just with DCCP?
[19:53:40] <kohler> don't know mobile sockets. probably the answer is yes. :)
[19:53:43] --- jlcjohn has joined
[19:53:47] <kohler> that app-layer text is old
[19:53:50] <lars> same thing just for tcp
[19:54:04] <kohler> is that Snoeren's work?
[19:54:12] <lars> i think so, need to google it
[19:54:29] <kohler> so this is different because unreliability makes everything much much much much easier
[19:54:42] <kohler> you can just dump all data into the socket regardless of sequence
[19:54:56] <kohler> snoeren's work, which i do know, has a lot of sequence number worries
[19:55:06] <lars> ok, but why is that an argument for doing it at the transport layer?
[19:55:06] <kohler> also, with a smaller option space, harder to authenticate
[19:55:12] <kohler> but yes, definitely similar
[19:55:24] <kohler> snoeren's work is transport layer
[19:55:36] <magnus> Eddi, unreliability does not make things easier. RTP and jitterbuffer would still be problematic with changed used of gencons.
[19:55:45] <lars> implementation maybe, but conceptually?
[19:55:59] <kohler> too philosophical, would have to think :)
[19:56:59] <kohler> MAGNUS IS RIGHT (but i'm not sure that matters)
[19:57:18] <lars> want me to tell pasi? :-)
[19:57:40] <kohler> sure! it might not matter because it's not clear this draft will take away from the working group's focus.
[19:59:09] --- shikob has joined
[19:59:29] <kohler> al: cool
[19:59:56] <kohler> those applications won't use multiple connections in parallel
[20:01:50] <magnus> How do you select which is the best one to use?
[20:02:04] <kohler> send an Attach message on that connection
[20:02:09] <kohler> or you can send a Data packet
[20:02:41] <magnus> No, how do you determine which is best. Not the protocol mechanism to do it.
[20:02:54] <kohler> oohh
[20:03:01] <kohler> not specified.
[20:03:41] <kohler> ah, sally :)
[20:03:45] <magnus> A pretty important problem to optimize for a media connection to get the best properies.
[20:04:07] <kohler> agree. may be application specific no? low-latency high-bw vs. high-latency low-bw ...
[20:04:51] <magnus> Yes, this is application specific. However it is one more of these applications problems that we need to write a story for if DCCP is going to be used.
[20:04:59] <kohler> oops, obviously that particular choice is easy.
[20:05:02] <lars> obviously, we need diffserv inside dccp :-)
[20:05:07] <kohler> yay
[20:05:25] <kohler> magnus: i don't know whether we need to solve those problems just to get the mechanism specified?
[20:06:20] <kohler> pasi wants the mechanism, and the draft was kicking around. he may solve it for his app, if it is important there. you are right, i wouldn't want the wg to turn into a 'multihoming' wg. i wouldn't think it would
[20:06:55] <magnus> No. But on the question on amount of work that means. I am a bit concerned with how we best use our cycles.
[20:07:22] <csp> The group is not exactly bursting with drafts and running out of agenda time...
[20:07:28] <kohler> agree. for me, the thing is, i have protocol mechanism cycles, but not at this point the kinds of cycles that you need.
[20:07:37] <kohler> :(
[20:07:50] <kohler> "Look at the blue thing!"
[20:09:29] <magnus> There is the question of review cycles. But maybe that needs a sexy mechanism to attract the cycles.
[20:09:35] <sarolaht> well, my world does not collapse if "multiaddress-dccp" is not accepted for wg charter...
[20:09:47] <sarolaht> i just thought it might be useful...
[20:10:29] <csp> Not sure review cycles are an issue here... there's no rush with the draft
[20:10:34] <kohler> i am happy with whatever the wg decides... the current mechanism is the first one i actually like. i'm happy to write down the crypto. if it dies, oh well
[20:10:52] <sarolaht> i'm not in a rush, either
[20:11:03] <sarolaht> i'm more like interested to see how this would work out
[20:11:30] <magnus> So lets get a good story for it. Because you may need it when going to the IESG and IAB.
[20:11:33] <sarolaht> for sctp it seems to work, based on some initial real-time trials
[20:11:55] <sarolaht> *real-life* ;-)
[20:19:59] --- magnus has left
[20:20:53] --- magnus has joined
[20:27:09] <kohler> i was assuming that'd be me for a while
[20:27:25] <csp> tag you're it :-)
[20:27:30] <kohler> OH GODDDDDDDDDDDDDD NOOOOOOOOOOOOOOO fine.
[20:27:45] <kohler> :)
[20:28:45] <csp> that's all folks
[20:28:54] <kohler> yayayay! thanks all for your help.
[20:28:57] <lars> bye
[20:29:01] --- shikob has left
[20:29:03] --- csp has left
[20:29:04] <kohler> byebye!
[20:29:04] --- magnus has left
[20:29:06] --- kohler has left: Logged out
[20:29:07] --- lars has left: Logged out
[20:29:10] --- arifumi@2entwine.net has left
[20:29:43] --- sarolaht has left