[02:06:21] --- sbrim has joined
[02:06:25] --- sbrim has left
[07:41:28] --- fred has joined
[07:41:35] --- fred has left
[07:51:19] --- psavola has joined
[07:54:22] --- AWGY has joined
[07:58:52] --- fred has joined
[08:01:38] --- gih900 has joined
[08:02:58] --- andrewdmcgregor@jabber.psg.com has joined
[08:03:12] --- shep has joined
[08:03:12] --- csp has joined
[08:03:25] --- jlcjohn has joined
[08:03:44] --- cabo has joined
[08:04:05] --- tsavo_work@jabber.org/Meebo has joined
[08:05:59] --- fujiwara has joined
[08:06:15] --- iljitsch has joined
[08:06:53] --- iljitsch has left
[08:06:56] --- iljitsch has joined
[08:07:43] <gih900> Where are the presentations for this session? They don't appear to be loaded into the IETF proceedings :-(
[08:08:32] --- dthaler has joined
[08:08:37] --- shikob has joined
[08:08:46] --- rhe has joined
[08:09:20] --- tony1athome@jabber.org has joined
[08:09:31] --- jyg has joined
[08:09:39] <tony1athome@jabber.org> Flames to the ADs. ;-)
[08:09:45] <dthaler> no one jabber scribing?
[08:10:01] --- resnick has joined
[08:10:07] <tony1athome@jabber.org> I'll try
[08:10:08] --- bensons has joined
[08:10:21] <tony1athome@jabber.org> BFD: going to last call with base spec
[08:10:26] <tony1athome@jabber.org> and some adjacencies
[08:10:27] <dthaler> would like at least a heads up towards the end of the wg reports so those of us that have to be in other WGs at the beginning know when to head over...
[08:10:30] <tony1athome@jabber.org> MIB not ready yet
[08:10:52] <tony1athome@jabber.org> Adrian: CCAMP: 6 new rfcs
[08:11:02] <tony1athome@jabber.org> 4 new drafts
[08:11:17] <tony1athome@jabber.org> some things tied up with security
[08:11:38] <tony1athome@jabber.org> interdomain signalling in last call
[08:11:51] --- mtcarrasco has joined
[08:12:04] --- dwf has joined
[08:12:15] --- pk has joined
[08:12:22] <tony1athome@jabber.org> isis: 9 drafts to proposed standard
[08:12:44] --- csp has left
[08:12:45] <tony1athome@jabber.org> L1vpn: incomprehensible
[08:13:01] <tony1athome@jabber.org> manet: several docs to last call, one experiemental
[08:13:16] <tony1athome@jabber.org> autoconf wg has manet arch
[08:13:49] <tony1athome@jabber.org> MPLS: one new rfc
[08:13:53] <tony1athome@jabber.org> 4 approved
[08:14:01] --- jariarkko has joined
[08:14:02] <tony1athome@jabber.org> GMPLS process document approved
[08:14:24] <tony1athome@jabber.org> need a milestone update
[08:14:39] <tony1athome@jabber.org> security framework needs to move out
[08:15:04] <tony1athome@jabber.org> charter needs to be examined as to whether it includes the work
[08:15:10] --- atarashi has joined
[08:15:19] <tony1athome@jabber.org> OSPF: no single manet technique for optimization
[08:15:30] --- csp has joined
[08:15:31] <tony1athome@jabber.org> continued refinement there
[08:15:44] <tony1athome@jabber.org> v2mib finished
[08:15:53] <tony1athome@jabber.org> was 4 years old
[08:16:10] <tony1athome@jabber.org> multi-area adjacencies, capabilityes, MTR draft finished
[08:16:28] <tony1athome@jabber.org> new: more applications (GENAPP) pw3
[08:17:05] <tony1athome@jabber.org> Doing something on optical control
[08:17:17] <tony1athome@jabber.org> Need to recharter soon
[08:17:35] <tony1athome@jabber.org> PCE: JP Vasseur
[08:17:51] <tony1athome@jabber.org> Almost done with requirement protocol
[08:18:00] <tony1athome@jabber.org> 5 implementations
[08:18:05] --- ian.chakeres@gmail.com has joined
[08:18:10] <tony1athome@jabber.org> report next ietf
[08:18:29] <tony1athome@jabber.org> 3 mibs documented already
[08:18:36] --- csp has left
[08:18:53] <tony1athome@jabber.org> RPsec: not present
[08:19:20] <tony1athome@jabber.org> RTGWG: not meeting; not talking about GTSM, moving to proposed standard
[08:19:33] <tony1athome@jabber.org> need to last call LFAs soon
[08:19:49] <tony1athome@jabber.org> should meet in Chicago
[08:20:04] <tony1athome@jabber.org> No SIDR chair
[08:20:06] --- harald has joined
[08:20:12] <gih900> no I'm not 'here" i'm elsewhere
[08:20:20] <tony1athome@jabber.org> VRRP: no present
[08:20:42] <tony1athome@jabber.org> MPLS: first draft standard
[08:21:18] --- csp has joined
[08:21:44] <tony1athome@jabber.org> IDR: 3 RFCs approved, one in last call, one in RFC editor
[08:21:47] <tony1athome@jabber.org> please see web site
[08:21:59] <tony1athome@jabber.org> Looking for commentors on v2 MIB
[08:22:05] <tony1athome@jabber.org> Trying to close after 5 years
[08:22:16] --- magnus has joined
[08:23:01] <tony1athome@jabber.org> WG reports done
[08:23:22] <tony1athome@jabber.org> Ross is talking
[08:23:35] <dthaler> time for roap related stuff now?
[08:23:55] <tony1athome@jabber.org> He's sorta monologing on it now.
[08:24:04] <tony1athome@jabber.org> Not clear there will be a discussion yet.
[08:24:49] <tony1athome@jabber.org> PS to give Geoff's slides
[08:25:04] <tony1athome@jabber.org> Dave W to talk about improvements; then open discussion
[08:25:12] <jlcjohn> URL for slides??
[08:25:16] --- csp has left
[08:25:37] <tony1athome@jabber.org> Hello Geoff!
[08:26:08] <tony1athome@jabber.org> PS is an F1 fan. ;-)
[08:26:10] <gih900> coming - 1 sec
[08:26:21] <tony1athome@jabber.org> Inter-domain Routing Trends
[08:26:40] <tony1athome@jabber.org> agenda
[08:26:51] <tony1athome@jabber.org> data; observations; further study
[08:26:55] <gih900> http://ww.potaroo.net/presentations/2007-03-22-routingtrends.pdf
[08:27:09] <gih900> http://www.potaroo.net/presentations/2007-03-22-routingtrends.pdf
[08:27:16] <tony1athome@jabber.org> v4 stats
[08:27:18] --- magnus has left
[08:27:23] <tony1athome@jabber.org> +17%
[08:27:27] --- mtcarrasco has left
[08:27:32] <tony1athome@jabber.org> AS growth + 13%
[08:27:48] <tony1athome@jabber.org> Advertisement size is getting smaller?
[08:27:49] <jlcjohn> thx
[08:27:54] <tony1athome@jabber.org> s/?//
[08:28:07] <tony1athome@jabber.org> about 24k asns today
[08:28:30] <tony1athome@jabber.org> Addresses per AS getting smaller
[08:28:35] <tony1athome@jabber.org> AS path steady 3.4
[08:28:46] <tony1athome@jabber.org> Interconnection rising (2.6)
[08:29:03] <tony1athome@jabber.org> stats for 2006
[08:29:12] <tony1athome@jabber.org> 354,000 unique prefixes
[08:29:18] <tony1athome@jabber.org> 2.8 prefixues per sec
[08:29:24] <tony1athome@jabber.org> 1 withdrawl per second
[08:29:36] <tony1athome@jabber.org> 203,000 updated
[08:29:41] <tony1athome@jabber.org> 150,000 withdraws
[08:29:47] <tony1athome@jabber.org> prefixes per update 1.95
[08:30:06] <tony1athome@jabber.org> Some kind of route leak causing too many prefixes
[08:30:23] <tony1athome@jabber.org> How good is this data?
[08:30:25] <gih900> nothing abnormal - just the ebb and flow of more specifics!
[08:30:44] <tony1athome@jabber.org> Slightly insulting to Australia as being less connected
[08:30:50] <tony1athome@jabber.org> Noisy data
[08:30:57] <tony1athome@jabber.org> Heavily skewed
[08:30:57] <gih900> but true in any case!
[08:31:22] <tony1athome@jabber.org> CDF of updates by prefix
[08:31:38] <tony1athome@jabber.org> 10% of perps generate 60% of work
[08:31:54] <tony1athome@jabber.org> 60% of the prefix account for 10% of the updates
[08:32:05] <tony1athome@jabber.org> beacons
[08:32:10] <tony1athome@jabber.org> 1 hour cycle
[08:32:28] <tony1athome@jabber.org> 4380 beacon events in 2006
[08:32:48] <tony1athome@jabber.org> Recorded 55,634 updates, 4423 withdrawls
[08:32:55] <tony1athome@jabber.org> Much amplification
[08:32:58] <tony1athome@jabber.org> 11:1 ratio
[08:33:06] <tony1athome@jabber.org> CDF by origin AS
[08:33:22] <tony1athome@jabber.org> 3% of ASs wiere source of 60% of all updates
[08:33:36] <tony1athome@jabber.org> 10% of updates were from 75% of ases
[08:33:55] <tony1athome@jabber.org> Most folks are very stable, lotsa good citizens
[08:34:05] <tony1athome@jabber.org> AS 17974
[08:34:15] <tony1athome@jabber.org> 1.3M updates in 06
[08:34:36] <tony1athome@jabber.org> What's happening?
[08:34:57] <tony1athome@jabber.org> Multi-homeing & TE
[08:35:06] <tony1athome@jabber.org> all intentional
[08:35:17] <tony1athome@jabber.org> unstable configurations
[08:35:36] <tony1athome@jabber.org> Where to now?
[08:35:51] <tony1athome@jabber.org> Do we need new routing structure or just fix BGP?
[08:36:01] <tony1athome@jabber.org> More intelligent dampening?
[08:36:22] <tony1athome@jabber.org> More use of standard mechanisms for propoagting BGP?
[08:36:40] <tony1athome@jabber.org> Can we tweak withdrawls?
[08:36:59] <tony1athome@jabber.org> Changing BGP is hard
[08:37:14] <tony1athome@jabber.org> Turns as fast as an aircraft carrier
[08:37:25] <tony1athome@jabber.org> needs to be backward compatible; no drastic changes
[08:37:35] <tony1athome@jabber.org> 32-bit ASN takes time to deploy...
[08:37:55] <tony1athome@jabber.org> Time scale is long
[08:38:05] <tony1athome@jabber.org> More study
[08:38:16] <tony1athome@jabber.org> what is BGP really doing? Need more observation points
[08:38:46] <tony1athome@jabber.org> How do we want to measure BGP?
[08:38:57] <tony1athome@jabber.org> Did our changes hep?
[08:39:08] --- magnus has joined
[08:39:14] <tony1athome@jabber.org> What happens when we make changes? Model, simulate to understand impact?
[08:39:22] <tony1athome@jabber.org> done
[08:39:30] <tony1athome@jabber.org> No questions...
[08:39:37] <tony1athome@jabber.org> Scudder
[08:40:05] <gih900> where is John's presenrtation?
[08:40:26] <tony1athome@jabber.org> Electrons are still wet, I think.
[08:40:35] <tony1athome@jabber.org> These are opinions of the authors
[08:41:07] <tony1athome@jabber.org> authors: scudder and ward
[08:41:11] <tony1athome@jabber.org> agenda
[08:41:16] <tony1athome@jabber.org> goals and priorites
[08:41:23] <tony1athome@jabber.org> goals: maximize connectity
[08:41:36] <tony1athome@jabber.org> convergence and stability are subsidiary to connectivity
[08:41:57] <tony1athome@jabber.org> Priorities: first: fast service restoration
[08:42:08] <tony1athome@jabber.org> Second: minimize peak load on control plane
[08:42:56] <tony1athome@jabber.org> focus on performance and stability
[08:43:16] <tony1athome@jabber.org> there are other aspects that are not covered (services, operations, wedgies, secruity)
[08:43:34] --- narten has joined
[08:43:45] <tony1athome@jabber.org> ASes are for loop suppression -- and nothing else (well, not really true)
[08:44:01] <tony1athome@jabber.org> ASes are NOT locators No topological signficants
[08:44:13] --- magnus has left
[08:44:14] <gih900> ASes are for loop supression and path metric
[08:44:26] <tony1athome@jabber.org> exactly.
[08:44:44] <tony1athome@jabber.org> auto-aggregation is a non-starter
[08:44:50] <tony1athome@jabber.org> manual proxy aggregation is too tricky
[08:45:05] <tony1athome@jabber.org> BGP is now multiprotoco
[08:45:16] <tony1athome@jabber.org> not all AFs need to be on all routers
[08:45:29] <tony1athome@jabber.org> VPNs
[08:45:37] <tony1athome@jabber.org> VPNS larger than Internet table
[08:45:48] <tony1athome@jabber.org> True in aggregate, not true of any singe VPN table
[08:45:54] <tony1athome@jabber.org> [Not sure I believe that yet]
[08:46:05] <tony1athome@jabber.org> Inherently paralelizatble
[08:46:14] <tony1athome@jabber.org> No single PE or RR holds ALL VPN tables
[08:46:27] <tony1athome@jabber.org> Operational chllenges in management
[08:46:36] <tony1athome@jabber.org> Some tools to do this already
[08:46:40] <tony1athome@jabber.org> Dynamic behavior
[08:46:49] <tony1athome@jabber.org> Confusion among routing experts
[08:46:58] <tony1athome@jabber.org> Surprising behaviors are possible
[08:47:06] <tony1athome@jabber.org> Important to understand bounding conditions
[08:47:10] <tony1athome@jabber.org> BGP runs over TCP
[08:47:19] <tony1athome@jabber.org> Flow control implys dynamics
[08:47:27] <tony1athome@jabber.org> Intuition about TCP is usually wrong...
[08:47:50] <tony1athome@jabber.org> BGP under load
[08:48:09] <tony1athome@jabber.org> W/O congestion, BGP passes updates as fast as possible
[08:48:16] <tony1athome@jabber.org> modulo MRAI, dampening
[08:48:28] <tony1athome@jabber.org> Degration mode under CPU results in state compression
[08:48:33] --- harald has left
[08:48:38] <tony1athome@jabber.org> "Adaptive low-pass filter" behavior emerges
[08:48:48] <tony1athome@jabber.org> Thisngs slow down, typically do not melt
[08:49:14] --- anabolism@jabber.org has joined
[08:49:51] <tony1athome@jabber.org> state compression: when router is slow, it gets data relatively slowly
[08:50:18] <tony1athome@jabber.org> getting sow data is not necessarily bad, it'll be more optimal when we get it.
[08:50:34] <tony1athome@jabber.org> Data size is always bounded
[08:50:43] <tony1athome@jabber.org> BGP adapts to speed of peer
[08:50:53] <tony1athome@jabber.org> One slow peer does not hinder convergence
[08:51:12] <tony1athome@jabber.org> Update packing: low prefix/update rations when not congested: that's fine
[08:51:22] <tony1athome@jabber.org> Higher rations under congestion ... which is needed
[08:51:42] <tony1athome@jabber.org> Packing ratios are scary
[08:52:19] <tony1athome@jabber.org> But packing is irrelevant during non-congestion
[08:52:23] <tony1athome@jabber.org> Convergence
[08:52:30] <tony1athome@jabber.org> at least O(n) in the size of the table
[08:52:40] <tony1athome@jabber.org> Fundametla to how the protocol works
[08:52:51] <tony1athome@jabber.org> Full convergence don't happen often
[08:53:10] <tony1athome@jabber.org> Hard to "fix" completely, but is it really broken?
[08:54:17] <tony1athome@jabber.org> Q (IVB):
[08:54:33] <tony1athome@jabber.org> No metric
[08:54:36] --- arifumi has joined
[08:54:43] --- cgn has joined
[08:54:48] <tony1athome@jabber.org> Doesn't work very well as network gets flatter
[08:54:58] <tony1athome@jabber.org> We can fix
[08:55:10] <tony1athome@jabber.org> JS: could, but won't help convergence
[08:55:20] <tony1athome@jabber.org> IVB: only fix convergence?
[08:55:23] <tony1athome@jabber.org> JS: no, more sides
[08:55:43] <tony1athome@jabber.org> Techniques to avoid full convergence: graceful restart
[08:55:45] <tony1athome@jabber.org> nonstop routing
[08:55:50] <tony1athome@jabber.org> fast reroute
[08:55:56] <tony1athome@jabber.org> pre-convergence
[08:56:14] <tony1athome@jabber.org> best-external, multi-path and similar hacks
[08:56:38] <tony1athome@jabber.org> Reflection
[08:56:46] <tony1athome@jabber.org> RRs hide backup paths
[08:56:49] <tony1athome@jabber.org> reduces RIB sizes
[08:56:55] <tony1athome@jabber.org> bad for convergence
[08:57:10] <tony1athome@jabber.org> Convergence: state reduction/data hiding
[08:57:15] <tony1athome@jabber.org> Faster convergence
[08:57:17] <tony1athome@jabber.org> Pick oone
[08:57:30] <tony1athome@jabber.org> RRs give sane management of BGP mesh
[08:57:36] <tony1athome@jabber.org> Centralizes management
[08:58:22] <tony1athome@jabber.org> RRs drop backup paths, that hurts convergence
[08:58:43] --- narten has left
[08:58:58] <tony1athome@jabber.org> Not covering why they aren't that good at data reduction
[08:59:04] <tony1athome@jabber.org> Known issues:
[08:59:07] <tony1athome@jabber.org> path hunting
[08:59:12] <tony1athome@jabber.org> nonconverging policies
[08:59:17] <tony1athome@jabber.org> O(n) in DFZ size
[08:59:24] --- cgn has left
[08:59:45] <tony1athome@jabber.org> Path hunting
[08:59:53] <tony1athome@jabber.org> well-known amplification effect
[09:00:04] <gih900> yes its path hunting and this is susceptible to local mods to BGP using a form of micro-damping
[09:00:14] <tony1athome@jabber.org> To reduce it: advertise root cause, propagate backup paths
[09:00:33] <tony1athome@jabber.org> Agreed
[09:01:02] <tony1athome@jabber.org> Backup paths
[09:01:14] <tony1athome@jabber.org> Transit ASes seldom fully partition
[09:01:27] <tony1athome@jabber.org> losing a singe inter-domain link you lose some routes
[09:01:36] <tony1athome@jabber.org> This looks like a blackhole
[09:01:50] <tony1athome@jabber.org> Due to data hiding by other routers and RRs
[09:02:29] --- glozano has joined
[09:03:05] <tony1athome@jabber.org> Speculation: does this cause amplification?
[09:03:17] <tony1athome@jabber.org> backup path propgation is featible
[09:03:21] <tony1athome@jabber.org> increases RIB state
[09:03:27] <tony1athome@jabber.org> Improves convergence tand stability
[09:03:58] <tony1athome@jabber.org> Tools
[09:04:01] <tony1athome@jabber.org> as-pathlimit
[09:04:04] <tony1athome@jabber.org> aggregate withdraw
[09:04:07] <tony1athome@jabber.org> best-eternal
[09:04:15] <tony1athome@jabber.org> best instrumentation reusing WRD infra
[09:04:18] <tony1athome@jabber.org> BGP free core
[09:04:25] <tony1athome@jabber.org> Dampening with better parameters
[09:04:27] <tony1athome@jabber.org> multi-path
[09:04:32] <tony1athome@jabber.org> root cause nofitication
[09:04:47] <tony1athome@jabber.org> What else is missing? This is a first pass?
[09:05:40] <tony1athome@jabber.org> [Summarizing each of these]
[09:08:21] <shep> by "BGP free core" does he mean MPLS within an ISP, or is he imagining a core of the Internet that is BGP-free (using what? MPLS?)
[09:08:35] <tony1athome@jabber.org> He says "pick your encap"
[09:08:47] <tony1athome@jabber.org> Presumably that includes MPLS
[09:09:01] <tony1athome@jabber.org> Which tools do we really need?
[09:09:17] <tony1athome@jabber.org> align costs & benefits to get deployment
[09:09:21] <shep> but I wonder whether he was thinking that inter-ISP would still always be done with BGP, or whether there would be something else there.
[09:09:35] <tony1athome@jabber.org> Typically it's still BGP
[09:09:40] <gih900> its still BGP as we know it
[09:10:05] <shep> OK, so by "BGP free core" he meant within a single ISP.
[09:10:08] <tony1athome@jabber.org> Yah, just in fewer places
[09:10:11] <tony1athome@jabber.org> Yah
[09:10:25] <tony1athome@jabber.org> Focus on behavior under load or making load go away
[09:11:18] <tony1athome@jabber.org> Dampening
[09:11:23] <tony1athome@jabber.org> misused in past
[09:11:27] <tony1athome@jabber.org> wrong default parameters
[09:11:48] <tony1athome@jabber.org> Very generous parameters only penalize worst flappers
[09:12:16] <tony1athome@jabber.org> 4 flap penalty is silly if you have 11:1 amplification
[09:12:30] <gih900> this is a withdrawl flap
[09:12:35] <tony1athome@jabber.org> What are right parameters?
[09:12:46] <tony1athome@jabber.org> Withdrawal flap still counts...
[09:12:49] <gih900> The key feature is that successive updates keep on increasing the AS path length then the prefix withdraws
[09:13:13] <gih900> and the update rate is based on the min route advert timer
[09:13:30] <tony1athome@jabber.org> If we have correct parameters, then this could be better
[09:13:34] <tony1athome@jabber.org> Punch line
[09:13:40] <tony1athome@jabber.org> BGP not in danger of falling over
[09:13:46] <tony1athome@jabber.org> [just slowin down]
[09:13:55] <tony1athome@jabber.org> IDR for near tierm improvements
[09:14:01] <tony1athome@jabber.org> RRG for fundamental changes
[09:14:08] <gih900> so one approach is to damp for (about) 120 seconds when you see successive attempts to lengthen the AS path on a prefix
[09:14:40] <tony1athome@jabber.org> [Many different approaches... to be sure... what do we implement?]
[09:15:15] <tony1athome@jabber.org> DF: should BGP be used as ID/LOC push mechanism
[09:15:18] <tony1athome@jabber.org> JS: Consider it
[09:15:31] <tony1athome@jabber.org> Need more data
[09:15:46] --- townsley@jabber.psg.com has joined
[09:15:52] <tony1athome@jabber.org> DF: Push model would be good
[09:16:20] <tony1athome@jabber.org> not everyone would need full table; do we build a separate BGP topology for mapping?
[09:16:25] <tony1athome@jabber.org> JS: Ok, maybe
[09:16:36] <tony1athome@jabber.org> DF: Should it be localized
[09:16:41] <tony1athome@jabber.org> JS: Yes
[09:17:01] <tony1athome@jabber.org> SH: What about confederations?
[09:17:14] <shep> Dino---- are you here in this jabber room? Do you have a jabber ID?
[09:17:22] <tony1athome@jabber.org> JS: Much of RR stuff is also true about confeds
[09:17:25] --- sbrim has joined
[09:17:48] <tony1athome@jabber.org> DMP: More about intra-domain routing
[09:17:49] --- fparent@jabber.org has joined
[09:18:03] <tony1athome@jabber.org> 5 million paths
[09:18:11] <tony1athome@jabber.org> this is the pain point
[09:18:32] <tony1athome@jabber.org> Tons of duplication due to RR
[09:19:09] <tony1athome@jabber.org> The problem is being ignored cause it's not sexy
[09:19:35] <tony1athome@jabber.org> JS: Reflectors look nice, but may make RIB bigger
[09:19:49] <tony1athome@jabber.org> DMP: 10-30x amplification
[09:20:09] <tony1athome@jabber.org> JS: Flat mesh might well be better
[09:20:47] --- ian.chakeres@gmail.com has left: Logged out
[09:20:48] <tony1athome@jabber.org> Yes, it's worth being looked at
[09:21:21] <tony1athome@jabber.org> DMP: add hierarchy with reflection
[09:21:54] <tony1athome@jabber.org> Robert Raszuk: If BGP is mapping, why not use AS as locator?
[09:21:59] --- andrewdmcgregor@jabber.psg.com has left
[09:22:06] <tony1athome@jabber.org> If you use this model, no new development is required
[09:22:25] <tony1athome@jabber.org> JS: AS shouldn't appear in packet headers
[09:22:42] <tony1athome@jabber.org> RR: I didn't mean they should appear in headers...
[09:22:51] <tony1athome@jabber.org> You should convert them to 4 byte addresses
[09:23:16] <tony1athome@jabber.org> Q: What is theoretical limit for BGP?
[09:23:26] <tony1athome@jabber.org> JS: No limit.
[09:24:06] --- dwf has left
[09:24:08] <tony1athome@jabber.org> Brian Carpenter: Aligned solutions are important; consider also net benefit
[09:24:27] <tony1athome@jabber.org> Folks wont' deploy little benefit
[09:24:30] <tony1athome@jabber.org> JS: True
[09:24:46] <tony1athome@jabber.org> Acee Linden: How to evaluate which tools?
[09:25:11] --- sbrim has left
[09:25:12] --- atarashi has left: Replaced by new connection
[09:25:36] <tony1athome@jabber.org> Do you sacrifice number of bits?
[09:25:55] <tony1athome@jabber.org> There's a problem in OSPF
[09:25:57] <tony1athome@jabber.org> Not in BGP
[09:26:03] --- iljitsch has left
[09:26:09] <tony1athome@jabber.org> LZ: There are more tools
[09:26:25] <tony1athome@jabber.org> Where should they go? Grow?
[09:26:36] <tony1athome@jabber.org> Quantitative analysis could be done
[09:26:50] <tony1athome@jabber.org> JS: Good academic work
[09:27:12] <tony1athome@jabber.org> No communication between academia & engineering
[09:27:36] <tony1athome@jabber.org> We should pay more attention to the literature
[09:27:51] <tony1athome@jabber.org> DW: Some sorting in the tool list (to BC's comment)
[09:28:10] <tony1athome@jabber.org> Limiting the scope of dynamics in the network are high
[09:28:36] <tony1athome@jabber.org> Convergence fixes may grow state
[09:28:42] <tony1athome@jabber.org> Must be careful tradeoff
[09:28:56] <tony1athome@jabber.org> Rajiv Sari:
[09:29:32] <tony1athome@jabber.org> You've shown problems with backup paths and too many paths. Reconcile?
[09:29:44] <tony1athome@jabber.org> JS: I don't think too many paths are part of the problem other than state?
[09:30:01] <tony1athome@jabber.org> Danny was really pointing to that. Means our intuition is wrong.
[09:30:39] <tony1athome@jabber.org> Must be careful.
[09:30:46] <tony1athome@jabber.org> Simulation, prototyping...
[09:31:08] <tony1athome@jabber.org> We've been focused on keeping state down
[09:31:18] --- shep has left
[09:31:21] <tony1athome@jabber.org> Do we need to maximize stability? Shift to more state?
[09:31:29] --- shep has joined
[09:31:35] <tony1athome@jabber.org> Not obvious tradeoff
[09:31:53] <tony1athome@jabber.org> RS: Multipath selection, advertisement, or both
[09:31:59] <tony1athome@jabber.org> JS: Advertisement
[09:32:04] <tony1athome@jabber.org> Add-path and multiple-path
[09:32:49] <tony1athome@jabber.org> RC: now open mike
[09:33:14] <tony1athome@jabber.org> Q: Some flap due to errors
[09:33:28] <tony1athome@jabber.org> Better tools and education?
[09:33:40] --- atarashi has joined
[09:33:42] <tony1athome@jabber.org> Centraize all route computation?
[09:34:06] <tony1athome@jabber.org> Work with mgmt community?
[09:34:19] <tony1athome@jabber.org> Express policy thru mgmt
[09:34:27] <tony1athome@jabber.org> Keep policies coordinated
[09:34:36] <tony1athome@jabber.org> I don't know consequences...
[09:34:47] <tony1athome@jabber.org> Possible intermediate solution?
[09:35:05] <gih900> My speculation was that the operators didn't even know the effects of their actions, let alone figure out that they needed a better policy management tool
[09:35:06] <tony1athome@jabber.org> Get tools into routers is our business.
[09:35:33] <tony1athome@jabber.org> DW: Ruediger, you've done this...
[09:35:34] <gih900> Even one attendee at the workshop was amazed to learn that his team was listed in the top 10 AS list
[09:35:48] <tony1athome@jabber.org> RV: I'm not ready, but...
[09:36:11] <tony1athome@jabber.org> A language that defines policies and services and be compiled into configurations would be useful.
[09:36:14] <tony1athome@jabber.org> Was RPSL
[09:36:26] <tony1athome@jabber.org> Doesn't do what I want, and it can be done
[09:36:41] <tony1athome@jabber.org> It would be useful.
[09:37:06] <tony1athome@jabber.org> DW: that could happen in grow. THey need new charter items
[09:37:30] <tony1athome@jabber.org> JA: Looking at test data, some of it was airlines, now trying to fix this in NEMO.
[09:37:42] <gih900> of the attendee was from the "airline industry" as Jari said
[09:39:01] <tony1athome@jabber.org> RC: we should do bgp work in IDR
[09:39:06] <tony1athome@jabber.org> SH: Yes master
[09:39:19] <tony1athome@jabber.org> DW: Bring me a latte
[09:39:25] <tony1athome@jabber.org> SH: I don't do windows either
[09:42:16] --- cabo has left
[09:42:35] --- bensons has left
[09:42:37] <tony1athome@jabber.org> What do we do where....
[09:42:43] <tony1athome@jabber.org> GROW seems obvious
[09:44:33] --- gih900 has left
[09:44:37] --- fparent@jabber.org has left
[09:44:40] --- tony1athome@jabber.org has left
[09:44:43] --- atarashi has left
[09:44:45] --- rhe has left
[09:44:48] --- resnick has left
[09:44:54] --- arifumi has left
[09:45:00] --- jyg has left
[09:45:11] --- shikob has left
[09:45:46] --- anabolism@jabber.org has left
[09:45:51] --- townsley@jabber.psg.com has left
[09:46:49] --- tsavo_work@jabber.org/Meebo has left
[09:51:31] --- glozano has left
[10:01:17] --- shep has left: Logged out
[10:02:29] --- jpc has joined
[10:10:48] --- jlcjohn has left
[10:12:15] --- dthaler has left
[10:15:22] --- AWGY has left
[10:26:52] --- psavola has left
[10:28:09] --- pk has left
[10:28:58] --- pk has joined
[10:43:34] --- fred has left
[11:07:32] --- arifumi has joined
[11:15:23] --- arifumi has left
[11:17:39] --- arifumi has joined
[11:28:25] --- pk has left
[11:38:12] --- arifumi has left
[12:09:18] --- jariarkko has left
[12:14:50] --- pk has joined
[12:14:51] --- pk has left
[13:46:25] --- jpc has left
[15:07:33] --- fujiwara has left