- Tuesday, November 5, 2013

mptcp@jabber.ietf.org

Tuesday, November 5, 2013< ^ >

ayourtch has set the subject to: Materials at https://datatracker.ietf.org/meeting/75/materials.html

Room Configuration

Room Occupants

GMT+0
[01:43:40] jlcJohn joins the room
[01:44:59] acmacm joins the room
[01:46:18] <acmacm> acmacm is jabber scribe
[01:46:39] <jlcJohn> :^)
[01:46:48] <acmacm> Chairs deck
[01:47:02] martin.stiemerling joins the room
[01:47:02] <acmacm> slide 4 milestones
[01:47:07] teco.boot joins the room
[01:47:36] <acmacm> slide 5 Agenda
[01:48:15] <acmacm> two sessions...
[01:48:53] <acmacm> security discussion is open ended tonight
[01:49:05] <acmacm> slide 6 News
[01:49:40] <acmacm> remote folks - can you hear ok
[01:49:45] <acmacm> ?
[01:50:27] rscheff joins the room
[01:50:49] Dan Wing joins the room
[01:53:54] <acmacm> Thomas Narten - odd that home router uses mptcp
[01:56:10] <acmacm> Phil can use multiple access links, yours, your neighbors
[01:56:33] <acmacm> Martin S. Security has not appointed an Advisor
[01:56:48] <acmacm> slide 7
[01:57:38] <acmacm> Summary of Interim meeting on security
[02:00:42] Dan Wing leaves the room
[02:03:59] <acmacm> Lars Eggert - what's the status?  consensus at interim and shared on list with no comment
[02:04:51] <acmacm> Lars - how do you avoid a downgrade attack?
[02:05:19] Dan Wing joins the room
[02:05:33] <acmacm> Marcelo - downgrade attack is well known, when you have mature deployment of prong 2 you can no longer accept prong 1 connections...
[02:06:08] <acmacm> Lars - going for record on most questions per minute
[02:06:30] koduck joins the room
[02:08:05] <acmacm> how to align use of TCPcrypt with regular TCP
[02:08:44] <acmacm> Marcelo - we can use any solution, need to have the alignment discussion
[02:14:57] sftcd joins the room
[02:20:01] <acmacm> summary of long discussion on encryption and keying: MPTCP always encrypts, it's not up to the app to choose
[02:22:30] <acmacm> but if you require encryption, you must use TLS, because a MPTCP connection could fall back to unencrypted TCP
[02:31:04] <acmacm> Laziness reduces the use of TLS
[02:32:40] rscheff leaves the room
[02:33:06] <acmacm> Phil - prong 1 can use more secure keys that you got somehow...
[02:33:42] <acmacm> very tentative favoring of TCPcrpyt (last line of slide)
[02:35:28] rscheff joins the room
[02:39:21] <acmacm> Calling Consensus questions on slide 8, question 1???
[02:41:47] <acmacm> call delayed - more discussion
[02:46:50] <acmacm> Need to work out the attack scenario, because may be covered, but maybe not
[02:47:23] <acmacm> Q1 should we go for 2 pronged approach?
[02:47:59] <acmacm> mild hum in the room for 2 pronged, silence opposed
[02:49:03] <acmacm> Q2:  prong 1 is to mitigate the ADD-ADDR attack, signal upgraded security, and Michael's addition
[02:49:19] <acmacm> slience for and against!
[02:51:01] <acmacm> Michael's addition to prong 1 on slide 7 is make sure that shared keys are available, then you can use it as part of HMAC
[02:55:32] <acmacm> No one seems to be strongly objecting, but it's hard to agree with these vague descriptions (?)
[02:56:28] <acmacm> skide 8 shoudl -mptcp-attacks be a wg doc?
[02:56:44] <acmacm> hum for, nobody against
[02:57:14] <acmacm> Alan FOrd on 6824bis
[02:58:55] <acmacm> Only doing ADD_ADDR, now 2
[02:59:22] <acmacm> as secure as MP join
[03:04:36] Dan Wing leaves the room
[03:04:53] <acmacm> MPTCP path selection using Port Control Protocol (PCP)  (15 mins)
     Dan Wing
[03:06:54] <acmacm> PCP extended to communicate the throughput delay and jitter that an APp would like to see.
[03:07:09] <acmacm> FLOWDATA
[03:08:10] <acmacm> Just communicate the MAXIMUM for a link, actual will aways be less
[03:20:47] <acmacm> Lots of questions about value for the fixed maximum of info on links
[03:21:41] Dan Wing joins the room
[03:21:49] <acmacm> If we could get reliable link info, capacity etc. we could certainly find uses for it. (LArs)
[03:22:30] <acmacm> MArcelo on connection acrobatics
[03:23:00] <acmacm> new use cases
[03:25:01] <acmacm> Host mobility provided by MPTCP and Intra Datacenter for load balancing and pooling links
[03:25:18] <acmacm> these are the new ones
[03:26:43] <acmacm> Waypoint Migration - move to alternate middlebox
[03:35:03] sftcd leaves the room
[03:35:47] <acmacm> Endpoint Migration
[03:36:30] <acmacm> - move to another server when no app state yet...
[03:40:07] <acmacm> Process migration, spin up a redundant VM in other center/city
[03:41:21] Dan Wing leaves the room
[03:42:34] Dan Wing joins the room
[03:42:40] <acmacm> Lars - having the client involved adds to latency, but Dan seemed to solve the issue
[03:43:16] <acmacm> DOes this work with prong 1 security? several people think yes
[03:43:37] Dan Wing leaves the room
[03:45:47] koduck leaves the room
[03:46:00] <acmacm> we're done!
[03:46:04] acmacm leaves the room
[03:46:06] teco.boot leaves the room
[03:53:24] rscheff leaves the room: Computer went to sleep
[04:01:46] martin.stiemerling leaves the room
[08:16:03] jlcJohn leaves the room
[10:23:07] rscheff joins the room
[10:24:58] rscheff leaves the room: Replaced by new connection
[10:24:58] rscheff joins the room
[10:47:57] rscheff leaves the room: Replaced by new connection
[10:48:01] rscheff joins the room
[10:50:43] rscheff leaves the room: Replaced by new connection
[10:50:43] rscheff joins the room
[11:29:46] rscheff leaves the room: Replaced by new connection
[11:29:46] rscheff joins the room
[11:30:33] rscheff leaves the room: Replaced by new connection
[11:30:33] rscheff joins the room
[12:36:12] rscheff leaves the room: Replaced by new connection
[12:36:12] rscheff joins the room
[12:38:11] rscheff leaves the room: Replaced by new connection
[12:38:11] rscheff joins the room
[12:39:24] rscheff leaves the room: Computer went to sleep
[15:21:46] sftcd joins the room
[15:48:03] martin.stiemerling joins the room
[15:48:25] martin.stiemerling leaves the room
[15:59:08] sftcd leaves the room
[16:02:08] rscheff joins the room
[16:05:10] rscheff leaves the room: Replaced by new connection
[16:05:10] rscheff joins the room
[16:29:02] rscheff leaves the room: Computer went to sleep
[16:52:06] rscheff joins the room
[17:17:39] rscheff leaves the room
[17:29:44] Dan Wing joins the room
[19:29:08] Dan Wing leaves the room
[21:11:26] Dan Wing joins the room
[21:39:31] Dan Wing leaves the room