[13:41:59] --- rikwade has joined
[13:47:58] --- rikwade has left
[13:47:58] --- rik wade has joined
[14:32:16] --- ggm has joined
[14:44:40] --- rpayn422 has joined
[14:45:59] --- javier has joined
[14:50:52] <ggm> Yakov kicks off. reviewing document status
[14:53:49] <ggm> Enke presents draft on dynamic capabilities draft-ietf-idr-dynamic-cap-05.txt
[14:55:40] --- yushun has joined
[14:56:01] <ggm> looking for mechanism to support dynamic exchange of capabilities like 4-byte AS, multipaths, which need ACK.
[14:56:33] <ggm> added ACK related fields in CAPABILITY message
[14:59:31] <ggm> specify ACK procedures. not backwards compatible. WG needs to choose method to deal with this. either use allocated capability , or get new one, deprecate old.
[15:00:44] <ggm> Yakov. don't use same number. semantics/coding different. its a big number space. lets use it.
[15:02:11] <ggm> [unknown on floor] change now is good.
[15:03:42] <ggm> detailed discussion on draft. encoding hacks, how to use 1 octet to flag future modes etc
[15:03:43] --- sakai has joined
[15:04:04] --- peterd has joined
[15:04:05] <ggm> enke: do we need more than 255 to encode the capability.
[15:04:12] <ggm> Yakov: take to Mailing List
[15:04:30] <ggm> draft-chen-bgp-group-path-update-01.txt
[15:04:34] <ggm> Enke again
[15:06:15] <ggm> discussing motivations
[15:09:03] <ggm> I can't keep up with this topic. its dense. meaningful jabber logging isn't going to be done by me.
[15:12:34] <ggm> comparing approaches. current is to adv. the best path, two alternates are to adv. the group best paths, all paths adv. with ewual to neighbour-AS count. or, adv, the group multi-paths. goal is to limit oscillation, routing inconsistency
[15:13:38] <ggm> How to encode the route withdraw in NLRI
[15:17:19] <ggm> implementation/deployment. only minor changes required for incremental deployment. need to select an approach. group best path is a good compromise. group multi-paths can involve too many paths. may need MED
[15:19:46] --- jis has joined
[15:22:23] --- rik wade has left: Replaced by new connection
[15:22:24] --- rik wade has joined
[15:25:29] <ggm> discussions about design. want ways to handle multiple paths more generally
[15:26:27] <ggm> graceful shutdown of BGP draft.
[15:28:31] <ggm> suggest inform peers first, then wait for maintenance timer. (manage FIB/RIB) -give peers time to re-work and converge.
[15:29:00] <ggm> applicable to a router restart or BGP session. not applicable for iBGP (mesh issues) -complex topic
[15:30:58] <ggm> issues in current draft: timer val. 30-60s. draft proposed 300s (too conservative) - scope extension. need to clarify iBGP. NLRI learned from multiple eBGP
[15:31:15] --- javier has left: Disconnected
[15:31:21] <ggm> benefits: no loss of traffic, easier config of maintenence evnts, no change to peers, BGP. feedback welcome
[15:31:47] <ggm> Q. whats difference to orderly withdraw of prefixes?
[15:32:12] <ggm> A the ordering is important. withdraw happens first,
[15:32:18] <ggm> Q no changes to BGP state engine?
[15:32:23] <ggm> A yes. its local.
[15:32:27] <ggm> Q how tested? put in route map?
[15:32:29] --- javier has joined
[15:32:29] <ggm> A yes.
[15:33:19] <ggm> yakov does chair process stuff. All you jabber people have to RUSH, nay I say RUN over here and sign the bluesheet. form a queue at the door..
[15:33:50] <ggm> next is group coop route filtering capability. presentd by susan hares
[15:34:26] <ggm> ORF filters == whole bunch of ANDing. may not provide efficient processing/expression of policy.
[15:34:55] --- dusty45 has joined
[15:35:00] <ggm> simple soln to group policies, apply in order, apply groups before non-group. numerical order. defaults should be specified. use defaults. move from AND to group of AND.
[15:35:13] <ggm> diagram of groupID wrapper around ORF
[15:35:45] <ggm> uses. layer3 vpn. global routing.
[15:37:24] --- javier has left: Disconnected
[15:38:02] <ggm> Enke Questions what it accomplishes
[15:38:39] --- javier has joined
[15:38:39] <ggm> Susan gives different ORF ordering option to whats in the standard
[15:39:43] <ggm> other Qestioner. what difference does processing the AND things make? why does I have to apply the order of the logical AND test change the result?
[15:40:10] <ggm> Susan. efficiency. lazy evaluation. secondly, some ORFs have both permit and deny. so it becomes an AND
[15:40:27] <ggm> Chandra. whose efficiency? the router doing it? or the one who is receiving it.
[15:40:30] <ggm> Susan the other end.
[15:40:38] <ggm> Chandra how can you pre-suppose whats efficient at the other end?
[15:45:26] <ggm> [missing the dialogue having to do other stuff in other windows. sorry. -ggm]
[15:47:01] <ggm> now on ATM reachability info in BGP
[15:47:18] <ggm> draft-ck-bgp-atm-yaddayaddayadda. god, these slides change too fast
[15:53:31] <ggm> that one was all about MPLS type l2vpn stuff.
[15:55:58] --- vm has joined
[15:59:50] --- javier has left
[16:03:39] --- orange has joined
[16:04:15] --- orange has left
[16:15:34] --- jis has left
[16:21:58] --- javier has joined
[16:22:03] --- peterd has left
[16:23:19] --- yushun has left
[16:23:51] --- rik wade has left
[16:30:45] --- sanjaya has joined
[16:32:26] --- sanjaya has left
[16:32:36] --- dusty45 has left
[16:34:13] --- sakai has left
[16:36:38] --- javier has left
[16:37:07] --- rpayn422 has left: Logged out
[16:58:33] --- ggm has left
[17:06:28] --- vm has left: Disconnected