[20:43:40] --- yjs has joined
[20:43:49] <yjs> testing
[20:44:03] <yjs> will be changes to order of agenda
[20:44:27] <yjs> (due to parallel timing with TSV meeting)
[20:45:16] <yjs> draft-mohan at end (may be extended discussion)
[20:46:34] <yjs> Stewart giving status review
[20:46:47] <yjs> lots of RFCs !
[20:46:55] <yjs> several expired MIB drafts
[20:47:19] <yjs> sonet in IESG processing, but held up on IPsec for UDP/IP
[20:47:38] --- yjs has left
[20:48:01] --- yjs has joined
[20:48:24] <yjs> FC being handled for congestion
[20:48:42] <yjs> comments required for AII
[20:48:50] <yjs> MS-PW passed to mark
[20:49:05] <yjs> VCCV - extension of WG LC
[20:50:31] <yjs> Ronen to talk about FC
[20:51:39] --- RicP has joined
[20:52:00] <yjs> reviewing additions in 02 version - reliable congestion control mechanism
[20:52:04] <yjs> signaling TLVs
[20:52:38] --- kenyis has joined
[20:52:45] <yjs> congestion control - rate control (TFRC) and thruput, loss event measured by receiving indication using LAP-B
[20:53:29] <yjs> LAPB - inherited from FC over ATM - is reliable protocol
[20:53:30] --- lllmartini has joined
[20:54:14] <yjs> LAPB has 2 types of frames - I (information) and S (supervisory - 3 types)
[20:54:52] <yjs> S frame types - receiver ready, not ready, selective reject
[20:55:18] <yjs> describing parameters for selective retransmission
[20:55:49] <yjs> LAPB header after CW
[20:56:16] <lllmartini> this PW type will create huge flows in an MPLS PSN .... we should work an a way to make it multi-link friendly .
[20:56:18] <yjs> adds LAPB A bit to CW flags field
[20:56:51] <yjs> LAPB header described
[20:57:31] <yjs> describing signaling extensions
[20:59:00] <lllmartini> yjs : how many bits do you use for sequencing
[20:59:18] <lllmartini> mohan : just reusing existing method
[20:59:34] <yjs> that's ronen not mohan
[20:59:46] <lllmartini> Setwart : is 16 bit sequience number enought ?
[20:59:50] <yjs> YJS asked since its 16 bits and could be copied into CW too
[21:00:17] <lllmartini> ok - thanks fixed in the minute too
[21:00:51] <lllmartini> PW security : Yjs speaking now
[21:03:40] <lllmartini> yjs explaining how macsec works in ieee 802.1
[21:19:09] --- yjs has left: Lost connection
[21:41:32] --- yjs has joined
[21:41:40] <yjs> yjs is back
[21:42:15] <yjs> not mentioned - "dotting the i's" slides - will take to the list
[21:42:38] <yjs> Cao Wei talking about 802.1ah PWs
[21:43:15] <yjs> PBT (built over 1ah) is also important
[21:43:36] <yjs> Ethernet PW not fully supportive of 1ah
[21:44:08] <yjs> scenario - tunneling over MPLS to interconnect 1ah domains
[21:45:12] <yjs> 1ah has B-tag and I-tag - either or both can be used to identify the PW
[21:45:27] <yjs> I-tag good for MS-PW case
[21:56:16] <yjs> Bruce Davie talking about congestion for PWs
[21:56:16] --- yjs has left: Lost connection
[21:56:33] --- yjs has joined
[21:56:56] <yjs> setting up strawmen - why don't need congestion control ...
[21:58:06] <yjs> Yakov Rekhter - why solve for PWs - multicast also not congestion friendly
[21:58:15] <yjs> chairs - we work on PWs here, not other things
[21:58:28] <lllmartini> lot's of things not congestion friendly - like NFS ....
[21:58:37] <yjs> Bruce - not true that PWs run only over well-engineered networks, may be run over Internet
[21:59:25] <yjs> PWs are so important - so should trample other traffic - but what if almost all traffic is PWs?
[21:59:28] --- RicP has left
[22:00:16] <lllmartini> Yakov is being hostile to this work.
[22:01:11] <lllmartini> yakov: we cannot solve the whole problem , hence we should not try to solve the minor pw problem
[22:02:22] <lllmartini> vach: what is this internet yo uare talking about ?
[22:02:41] <lllmartini> is this not just as set of well engineered networks ?
[22:02:55] <lllmartini> Bruce : we should just turn off TCP congestion control
[22:03:01] <lllmartini> ;-)
[22:03:33] <lllmartini> Yjs : Pw is different from UDP
[22:03:54] <lllmartini> because UDP was done wrong , we cannot justify doing pw wrong
[22:04:38] <lllmartini> Danny defends Bruce
[22:04:52] <lllmartini> Finally no more bashing questions ....
[22:04:54] <lllmartini> we move on
[22:05:25] <yjs> primary problems - TDM PWs (CBR), packet PWs with inelastic traffic (e.g. MPEG-2)
[22:06:31] <yjs> goal - protecting the network from a new traffic type
[22:06:45] <yjs> running everything over TCP is not an option
[22:06:59] <yjs> (too much state)
[22:07:13] <yjs> BW efficiency, no ACks, etc
[22:08:33] <yjs> design choices - how to detect congestion, how to return info to sender, what to do when congestion detected, revertive or not
[22:09:14] <yjs> detecting - SNs (when there) , OAM packets, ECN/PCN
[22:09:32] <yjs> feedback egress to ingress - OAM message
[22:09:51] <yjs> TFRC (RFC3448)
[22:11:08] <yjs> summary - measurement, frequency of loss measurement, enforcement of rate, mandatory or not
[22:16:47] <yjs> Lars - I am the IESG person who wanted this, I realize that most of the time not needed
[22:17:28] <yjs> but happy to see this as will cause implementers to think about it
[22:18:25] <yjs> Luca on dynamically placed PWs using PCE
[22:19:02] <yjs> in a complex nework can not guarantee best path by using BGP
[22:21:00] <yjs> demonstrates that hard to determine optimal path across multiple AS'es
[22:21:27] <yjs> solution is to have a PCEs that intercommunicate
[22:22:41] <yjs> not every router is S-PE, so only don't need to flood info in IGP, rather PCE - S-PE communication
[22:23:07] <yjs> constraints - BW, delay, diversity
[22:23:25] <yjs> (diversity - for BU paths)
[22:27:13] <yjs> questions ? YJS asks what is being suggested
[22:27:30] <yjs> Luca - protocol for PCE to S-PE and other pieces
[22:27:45] <yjs> Ben NJ (BT) optimization problem is hard
[22:28:17] <yjs> Luca wants to talk about OAM message draft
[22:29:49] <yjs> AD required mandatory use of PW status - now seems problematic
[22:32:54] <yjs> now that we know how to coordinate between BFD/VCCV and PW status, can we advance the draft?
[22:33:11] <yjs> Stewart - take to list
[22:33:53] <yjs> Mustapha on multihop VCCV
[22:34:41] <yjs> were two suggested extensions
[22:35:47] <yjs> 1) use existing VCCV CW and dec TTL 2) new MS-VCCV CW
[22:36:34] <yjs> proposal - both methods may be required (based on existing deployments)
[22:36:49] <yjs> version 04 of segmented-pw will have text
[22:38:08] --- kenyis has left
[22:38:21] <lllmartini> ok now it's just us ....
[22:38:21] <yjs> signaling - method 1 uses VCCV parameter as now, but new CW needs type 4
[22:38:32] <lllmartini> no need to type anymore .....
[22:38:43] <yjs> so need new PWACH channel type
[22:38:52] <yjs> I type for posterity
[22:39:13] <yjs> (the logs stay around)
[22:39:14] <lllmartini> Matthew is doing an outstanding job
[22:40:01] <yjs> Don Oconnor wants a separate MS-VCCV draft
[22:41:27] <lllmartini> more documents == more overhead , I do not believe it is justified
[22:41:52] <lllmartini> this will be in segmented pw doc
[22:41:58] <yjs> although a short 2 page RFC can be useful
[22:42:11] <yjs> Stewart - PWE for T-MPLS
[22:42:34] <yjs> SG15 of ITU is working on MPLS transport networks
[22:43:09] <yjs> T-MPLS was liaised to IETF rather late in process, when it was too late to change
[22:43:56] <yjs> after 2 meetings and much discussion between IESG members and ITU there are still concerns
[22:44:46] <yjs> not clear whether T-MPLS is a subset of Ethernet PW over MPLS, or an "extended subset"
[22:45:08] <yjs> problem is that T-MPLS uses the same Ethertype, and if not compatible could be problematic
[22:45:30] <yjs> prior to IETF involvement there was no requirements doc - now there is
[22:45:49] <yjs> present draft shows that requirements can be met with present MPLS and PWs
[22:51:05] <yjs> Andy M - very late in day, but we are not chartered to do this now, there are other ways to get change control
[22:51:45] <lllmartini> huge line at the microphone
[22:52:36] <lllmartini> line is : Vach , Ben niven jekins , Dimitri , Monique , Dave
[23:17:33] --- yjs has left: Lost connection
[23:21:38] --- lllmartini has left