[10:35:09] thomas.dreibholz joins the room [12:10:31] sbarre joins the room [12:19:32] sbarre leaves the room [18:32:50] thomas.dreibholz leaves the room [20:25:40] ꈲ joins the room [21:57:15] ꈲ leaves the room [22:03:37] ꈲ joins the room [22:17:42] gigix73 joins the room [22:22:21] MPTCP joins the room [22:27:51] Session is going to start [22:28:08] If anybody has question for the mic please write "MIC:" [22:28:55] Slide on what is MPTCP [22:29:08] Andrew McGregor joins the room [22:29:11] admin slide [22:29:29] Note Well [22:29:35] Progress summary [22:30:15] There has been an interim meeting [22:30:22] behcet.sarikaya joins the room [22:30:29] 2 wg docs have been created since Hiroshima [22:30:47] Shane Amante joins the room [22:30:59] slide with schedule [22:31:22] MArch 2010: establish WG consensus on arch and high level design [22:33:08] Slide on MPTCP Implementers Workshop [22:33:24] July 24 2010 Maastricht [22:33:28] 2pm [22:33:42] organized by the WG but not a WG meeting [22:34:45] OS people are very welcome [22:35:47] help to organize the event is welcome [22:36:09] Agenda slide [22:36:43] today : Architecture and protocol and discussion on options vs payload [22:37:25] MPTCP Architecture presentation by Alan Ford [22:38:04] Lars joins the room [22:38:06] slide 2 diff from ietf 76 [22:38:16] slide 3 purpose of the draft [22:38:49] slide 4 high level design decisions [22:39:13] (consensus verified at Interim) [22:40:23] slide 5 New decisions at Interim [22:41:42] Scoot Brim: they will doc the issues [22:42:02] slide 6 other changes from previous version [22:42:37] slide 7: Questions? [22:43:37] Tim: read briefly the draft, what do you mean by middle boxes? [22:44:15] Yuri: data ack coming from different paths? [22:44:45] A: yes can come from different paths [22:44:59] Mark: all paths all the time [22:45:22] ???: mixed v4/v6 can you explain? [22:45:44] A; we can have one v4 subflow and one v6 subflow [22:46:40] Tim: now you have ack inside headers vs single ID last time, it's a good thing that open possibilities [22:46:42] gorryf joins the room [22:47:34] Mark: charter says we need doc in august but better to keep doc open until protocol specs are issued [22:47:53] Jana: support keep doc open [22:48:09] usefl if somebody review the doc [22:49:08] Presentation MPTCP Protocol [22:49:31] slide 2 minor changes from 02 [22:51:47] Mark at the mic [22:52:09] slide 3 Data ack [22:53:15] slide 4 middlebox issues [22:54:26] slide 5 remaining open issues [22:55:32] marcelo: can you expand on the policy issue? [22:55:58] A; send information about links that are most costly [22:56:16] Marcelo: what policy to open a subflow? [22:56:26] A; true need discussion [22:56:40] Mark : you think we need protocol change? [22:57:14] Marcelo: you can signal different things do know how [22:57:55] slide 6 any question? [22:58:32] Eric: impact of non unique IP addresses? [22:59:26] A: the token helps [23:00:07] Mark: there is another draft (not referenced) that proposes a solution [23:00:14] behcet.sarikaya leaves the room [23:00:39] S7365468EEE7F6 joins the room [23:00:49] Eric: there can be cases not covered and very difficult to fix [23:01:12] PasiS joins the room [23:02:07] Mark: there are cases where you do not advertise mp capabilities because you do not know what's on the other side [23:02:54] Erik: how to use multiple public addresses when I have one (or more) private addresses [23:04:24] Next presentation by Costin Riciu [23:04:58] Lars: anybody decided to implement? [23:05:08] except the trilogy linux based? [23:05:18] PasiS leaves the room [23:05:29] Costin: UCL belgium have an implementation [23:05:42] more stuff for the next IETF, may be a Demo [23:06:23] Mutipath TCP Signaling options vs payload [23:06:27] slide 2 motivation [23:06:34] S7365468EEE7F6 leaves the room [23:07:03] slide 3 what's at stake [23:07:43] slide 4 considerations [23:09:13] There is no evidence that the options you are talking about made it to the server [23:09:19] PasiS joins the room [23:09:40] Mark: is difficult to do the experiment, but the behavior is senseful even if you do not know [23:09:49] slide 5 options encoding [23:10:13] behcet.sarikaya joins the room [23:10:42] slide 6 payload encoding [23:11:34] Erik: on previous slide: how many middleboxes mess up with options and data packets? [23:12:32] C: If the syn +option passes also the data options should pass [23:13:17] Mark: middleboxes can drop for security [23:13:48] Scott: does not help to know if options pass, we are doing mptcp [23:14:28] C: if everyone deploys the new generation middleboxes will change accordingly [23:16:02] noname: medium term vision middleboxes vendors are not able to implement to let data options go through because they do not handle that [23:17:02] Anantha: touching data is dangerous somehow [23:17:47] Anatha: you are assuming data is also tlv [23:17:58] C; yes but you can have important gain [23:18:30] Anatha: the implementation can be tricky, middleboxes are not a problem IMO [23:18:38] tsavo_work@jabber.org/Meebo joins the room [23:19:46] Nishida: you cannot optimize MPTCP if middleboxes do not negotiate between them [23:21:44] Jana: middleboxes assume bytestream they can play with the sequence numbers and packets [23:22:29] Lars: third option: have a control subflow [23:22:52] Marcelo:what if that one goes down [23:23:07] Jana: signaling as payload of that flow [23:23:10] Lars: yes [23:23:19] C: is very tricky [23:23:38] slide 7 connection/subflow setup [23:24:53] slide 8 data sequence mapping [23:27:40] slide 9 Data ack [23:30:49] Jana: which ack is triggering more data? [23:30:57] C: subflows acks [23:31:56] Jana: when data ack are in the payload data ack just say you can drop the packet from your buffer no effect on congestion control which is done by tcp [23:32:04] c: correct [23:32:49] Jana: if buffers are large you have the same semantic as new reno [23:33:17] C: you can make your sender be less aggressive like reno, right [23:33:35] Why not send data acks as urgent data? [23:34:19] You want that I rely the question to the mic? [23:36:48] Yes, that would be good [23:37:10] if anybody has question for the mic please write "MIC:" [23:41:05] @andrew: Mark just told me they explored the issue, better if you send an email to him. [23:41:20] slide 9 security information [23:42:35] slide 10 middleboxes [23:42:57] slide 11 future middleboxes [23:44:34] slide 12 future middleboxes (2) [23:46:05] slide 13 single path MPTCP [23:46:39] slide 14 comparison summary [23:47:45] ruri joins the room [23:47:48] Alan: security you do not use payload for the whole connection [23:48:12] slide 15 opetions:what next [23:48:32] s/opetions/options [23:49:12] slide 16 Payload encoding: what next [23:50:12] slide 17: What should we standardize? [23:50:16] behcet.sarikaya leaves the room [23:52:37] Phil: any issue missing from the comparison table? [23:53:46] Erik: how to decide when you put tlv in the data, either you do it always or never [23:54:41] C: even if you receive garbage at the other end you drop because does not respect the expected format [23:56:12] C: there is an implementation working [23:56:32] Erik: not sufficient you need to guarantee it works always [23:58:50] Tim: middlebox sometime behave in such a way that touches the data [23:59:52] evyncke joins the room