INTRO
1.     Bill Fenner thanked for his previous work leading the group David Ward welcomed

2.     Reports from the individual working groups

McBride: PIM spec published, milestones completed, new charter agreed with PIM refresh reduction included.

Dave Ward , BFD:      Last call on base and corollary drafts. Point to multi-point added to charter

Adrian Farrel, CCAMP:  6 new RFCs 4 more in RFC editor queue. IESG have issues on 2 documents, but progress is being made to resolve those issues. Inter-domain signalling work in last call.

Dave Ward ISIS: 9 informational RFCs have gone to proposed standard. 8 drafts are intended for proposed standards track

Adrian, L1VPN:       Framework and requirements out, solutions under progress

Ian Chakeres, manet:         Several items in last call, there is one experimental protocol and some building blocks. draft-ietf-autoconf-manetarch-01 in autoconf group is also relevant.

Loa Andersson, MPLS:   1 new RFC, another 4 approved. Change process for MPLS and GMPLS approved. Good meeting. There are still things within the current charter that need doing. Ross Callon:   security framework - is this in the charter? Loa:    We need to look at the charter again

Acee OSPF Running with 3 experimental drafts. OSPFv2 MIB finished at last. Some more documents still to finish. More applications are using OSPF to disseminate information, this is leading to more work. Hope PWE3 doesn't want to use OSPF like this. Thinks there are L1 optical control plane issues to consider. Close to needing re-charter.

JP Vasseur, PCE Requirement docs for comms protocol getting quite stable. 5 implementations, interop. report expected for next IETF meeting. Worked with ?? on discovery. There are 3 MIB documents, on discovery and signalling and a manageability for each protocol document RPSEC -

John Scudder RTGWG There will be no meeting on Friday. GTSM last call and going for proposed standard in next few weeks. Last Call ?? Soon intend to take that to proposed standard. May need meeting next time as there is some new stuff coming along.

Loa Anderson SIDR First draft std coming through this week. needs to be evaluated again. would like to see if we could improve the process as they spent 2.5 years improving nits in proposed std txt.

Sue Hares, IDR. 3 produced, 2 approved, 3 last call, 3 RFC editor queue. Looking for BGPv2 MIB people to comments Want to close group down due to lack of activity, so if you care shout.

Ross Callon discussed the on relationship of different routing and addressing activities. regardless of the outcome of work on the Loc/ID split we still need to do routing. The split might not solve all our problems - table size may still grow so what can we do to make IDR work better. This is business as usual

3:     Peter Schoenmaker presented Geoff Hustons slide set on IDR routing trends

4:      John Scudder (JS) presented for himself and David Ward “Thoughts on Improving Inter-Domain Routing,” after giving disclaimer that what he was saying was his interpretation and not meant to be the official group opinion.
    Iljitsch van Beijnum:         You say you can't get better convergence with BGP than we currently have. We are now using AS path length as the inter-domain metric, but this is not good anymore as the network is getting flatter. can we not fix this?
    John Scudder:   you can fix this but it won't help convergence
   Dino Farinacci :      Do you think BGP should be used as an ID to loc push mechanism?
   John Scudder: If we want control driven mapping then BGP should be considered as it does work
   Dino Farinacci :      benefit is you don t have to drop packets or have routable ids...we may need BGP the protocol but with a different peering topology, using existing peering topology and moving mappings around might be bad as there could be stuff all over the core
   Sue Hares:      You don't mention confederations? AS confederations give some sort of scaling
   John Scudder:   What is true of route reflectors is also true of confederations
   Danny McPherson:      Surprised there is not more about intra-domain routing. Because of how route reflectors and policies work, I have a 30 times level of duplication on number of paths stored (8000 paths between devices, 5 million stored on router). I think the problem will hit ISP networks first, not the inter-domain
    John Scudder:   The route reflectors certainly may be making RIB bigger Need fewer more useful paths, need to use route reflectors as an operational tool only, we may need to redesign route reflectors to avoid route amplification, adding hierarchy can fix this but can lead to oscillations
    Robert Raszuck: I am surprised that you dismissed the use of AS# as locator? overloaded semantics is insufficient justification.
    John Scudder: I don't want to see AS# in packet header, I meant no more than that RR - could convert AS number into something in the address
   Jerome: Imagine we had infinite bandwidth, CPU, is there a fundamental limit at which BGP will stop?
   John Scudder : Not if everything is infinite, but we need to think about the finite limitations
   Brian Carpenter:      Solution deployment is very important. But we need to know how much of the pain any solution fixes - we don't want to make people install stuff  if its only minor benefit, avoid many upgrades Ans: need to maximise the bang for the buck
   Acee: How would you evaluate which tools you should use? would you need more control plane?
   John Scudder # bits BGP generates is trivial, we are a long way from being i/o bound - very different to OSPF
    Lixia Zhang:    there are many more possible fixes than listed already out there. I have sent a mail to the GROW mailing list. Quantitative analysis is required
    Scudder:    yes there is a lot of good stuff. There has been poor research to engineering down-streaming, we need to read what they have done
    David Ward:     We can structure the tools into categories, some are to control dynamics, and others are more operational. Some fixes for overall convergence property of BGP actually makes  databases larger John Scudder: we have a time memory trade off and need to understand where we should be
    Rajin Sari:     problem is that the number of paths in every router, then you say it helps to have a 2nd best path also stored, do these not contradict each other?
    John Scudder: Is the # of paths the problem? Danny pointed out earlier that our intuition on this is often wrong. We need simulations and experiments to determine if this would make things better or worse. There has been a bias in recent years to keep memory requirements down and we haven't thought enough about maximising stability. We may need more memory for better stability, the trade off is not clear.
    Rajin Sari:     were you talking multipath selection or advertisement
   Scudder: Advertisement
   Elwyn Davies:   Rose workshop showed that much churn was due to operator error. Some of the things we could do are not protocol change but tools to help people control their routers. We should cooperate with the management community to get interfaces to routers to enable things like centralised policies that could be downloaded to routers. Tools is not ietf business but how to deliver that info to the routers is
    Rudiger Volk:  level and language that can define policies to compile into router would be useful. RPSL was an attempt but too limited
    Dave Ward: where should the work take place? GROW is operational Requirements and looking for new charter?
    Jari A: Airline industry is causing a lot of load to IDR update so we have started work in the internet area to look at a way to support these airline routers without impacting the routing tables
    Ross Callon:    What aspects should we actually peruse? Do we extend the charter?
    Dave Ward:       Modifying AS# usage is architectural change, so belongs in RRG, as do new design patterns for BGP new network design patterns that allow operators - belongs to operator forums? Ross Callon: but there is no worldwide operator forum, shouldn't that be in GROW
    Dave Ward:      OK, Grow for policy interfaces & current scaling issues in building networks and design patterns to alleviate the problems
     Danny McP. Wasn't grow talking about closing? but stuff might be better in ops area?
    DWard - yes
    Dave Ward - we have SIDR for interaction with security but there is no interaction with apps and transport. Probably RRG should take into account those interests and ROAP
     Bill Fenner ops ADs want to close GROW as its been winding down so action required to keep it alive

6:      Close meeting