IDR 26 June 2007-07-27, 1:00 pm Minutes: Geoff Huston Agenda was bashed 1. Document Status: Change since IETF68 - 4-octet AS published: RFC4893 - Autonomous System Confederatuiobn for BGP - revised draft submitted to RFC Editor - Avoid BGP best path transitions from one external to another - moved to RFC Editor queue - New WG document - draft-ietf-idr-v4nlri-v6nh-00.txt 2. PWE3 - Dynamic placement of MS-PW (draft-ietf-pwe3-dynamic-ms-pw-05.txt) Introduced to IDR to solicit comment on how BGP is to used in this pwe3 context. [slides to illustrate the proposed mechanism] Routing information (reachability) and provisioning (signalling) is used in VPLS autodiscovery, while dynamic ms-pw uses only reachability as provisioning is static The BGP NLRI is a 96 bit quantity of 3 x 32 bit components (global-ID, prefix and AC-ID) Requesting early allocation process for allocation of BGP SAFI = 6 for this NLRI Questions/Comments: The WG Chair noted that this is a PWE3 WG document, and the chair of the PWE3 WG has the role to manage the early allocation process Question on the use of private AS numbers in the Global-ID field of the NLRI? Does it have to be an AS or any globally unique 32 bit value? Response is that any 32 bit unique value oculd be used here. Question put to the WG if there was any WG comment on the PWE3 WG early allocation action? No comments from the WG. 2. Flow-spec (draft-marques-idr-slow-spec-04.txt) [slides] Approach to convey network transport layer attributes with a given NLRI in BGP. Adopt this as a WG document? Is this a capabilities exchange? Yes, this is exchanged only if there is a capabilities exchange here, so it will / should not leak inadvertently beyond understood scope. Operational deployment exists. Chairs will poll the mailing list with the document adoption question. 4. BGP Encapsulation SAFI and BGP Tunnel Encapsulation attributed (draft-pmohapat-idr-info-safi-02.txt) [slides] Adoption as a WG document? Noted that this is intended to be a normative reference by the softwires mesh framework in the Softwires WG. This is a similar case to the v4nlri-v6nh document. Chairs will poll the mailing list with the document adoption question. 5. BGP Stability Improvements (draft-li-bgp-stability) [slides] Document is work in progress. Considers the model of BGP as an ideal of information dissemination through single wavefront, but BGP, as deployed, has differing propagation speeds in differing networks, leading to "diffraction" of this information wavefront that amplifies the original information event into multiple events. The document explores some approaches to improve the dynamic behaviour in BGP based on using additional information related to heuristics that attempt to identify "diffraction" in BGP information propagation. Tony called for collaborators - literature search, analysis, coding and testing EFT. Questions / Comments: John Scudder suggests a worked example for each of the suggested mechanisms as some of the description is not clear. Also suggests that the goal here is to maximize connectivity here. For example removal of the MRAI timer is consistent with this. Illjitsch van Beijnum commented that these suppression techniques are poor as they negatively impact some particular use cases. Sue Hares noted that some "crisis" examples (failure) where failures would trigger motivations for rapid repair rather than damped convergence Yakov asked for more information about the small proportion of prefixes that are the subject of large volumes of updates John Scuddr on best path latching and the combination of this with path exploration is interesting - what is the effect of best path latching in isolation and is this useful? Tony commented that by itself there are some strange interactions here that do not necessarily "do no harm: Yakov - section 4.3.1 question about peak rate per hour? Clarified as a peak per second rate, expressed in hourly intervals. Lixia noted that there is a large set of literature on BGP analysis, and there are many ideas in this topic in this literature.