Editor's Note: Minutes received 7/31} CURRENT_MEETING_REPORT_ Reported by Matt Mathis/PSC Minutes of the Border Gateway Protocol Deployment and Application Working Group (BGPDEPL) Executive Summary The Internet BGP topology was presented. There are about 25 administrative systems which are reachable from the NSFnet via EGP2, and about 40 which are reachable via BGP. Peter Lothberg described the European EBONE BGP deployment. In several ways EBONE is further along in BGP deployment than US networks. Several vendors described the status of their implementations. CISCO and Cornell gated have BGP3 running today, and are working on BGP4. BBN has BGP3 running and is working on BGP4. Other vendors including Proteon and Wellfleet are actively working on BGP3 in preparation for BGP4. ANS intends to have BGP4 deployed January 1993 and has offered to help vendors with interoperability testing. Vendors can arrange to bring equipment into the ANS test facility. ANS is also exploring support for remote testing by ``tunneling'' BGP from the ANS test network through the NSFnet. Contact Jordan Becker for further information. Minutes Two representations of the known BGP topology of the Internet were presented. The first was a map compiled by Jessica Yu, of Merit, which is compilation of several sources including Merit configuration databases and routing information extracted from the NSFnet T1 and T3 backbones. The other was BGP routing table as extracted from a router in Pittsburgh which was peering with both NSFnet backbones. The BGP topology includes about 40 administrative systems. There are only 25 administrative systems which are visible with EGP2 as the protocol of origin. Note that EGP2 does not propagate further topological detail, so there are additional EGP2 only administrative systems beyond the 25 which are visible. One interesting artifact was noted: The T1 backbone does not run IBGP so it can not propagate BGP path information. All BGP routes learned via the T1 backbone show one administrative system beyond the backbone itself with the ``incomplete'' origin attribute. This is dangerous because it completely defeats BGP loop suppression. Peter Lothberg described the European EBONE BGP deployment. They are using BGP3 in an unusual configuration where it is, in effect, their interior routing protocol. All EBONE routers are border routers. (There are no interior routers). The EBONE IGP (cisco's IGRP) is used 1 only to distribute the EBONE internal links and interfaces. The IGP does not carry exterior routes at all. IBGP is run as a fully interconnected mesh between all routers. Tony Li of cisco added a configuration feature to BGP to defeat the IGP alignment check. Thus every router has an IBGP route for for all exterior networks and an IGP route for EBONE routers. Packet switching takes two route lookups: one to select the exit router and then to find the interior route to that exit router. Peter described several routing misadventures that they encountered, as well as European configuration management issues and related politics. Several vendors described the status of their implementations. Nearly all of the major router vendors were present and actively working on various aspects of BGP. o Proteon plans to support BGP3 in release 13 for ``all platforms'' by the end of the year. o Wellfleet plans to have BGP3 in mid '93. o BBN will be replacing the MILNET mailbridges by T20s in about six months. the T20's will do IBGP with each other, and will support BGP2, BGP3 and EGP2. o cisco did a full rewrite of their BGP code earlier this year. Version 9.02 has many patches from early testing in Alternet and NEARnet. Check release notes on ftp.cisco.com. o The T3 backbone (ANS) will be running BGP3 in the fall, using a gated based implementation. (The current version is routed based). They plan to have BGP4 early '93. o CA*NET is currently using BGP3 in gated. BGP4 will require significant changes to gated, but may be available in the fall. The current Cornell gated code includes the CA*NET BGP3 code. There was discussion about what the mid-levels could to help the vendors. Tony Li of cisco indicated that test installations would be useful. Their primary tool has been core dumps from sites which are having problems. It was noted that it is difficult to test BGP in a laboratory, because the interesting behaviors come from its ability to represent complex events in the global internet. Jordan Becker volunteered to provide vendors with access to the ANS test facility for BGP interoperability testing. Contact him for details. There was also discussion about adding a feature to the ANS gated to support BGP tunneled from the ANS test facility across the Internet. This would permit remote vendors to do initial interoperability testing in their own labs. The necessary code change is trivial but there are 2 some operational and scheduling issues to be addressed. Contact Jordan Becker. There was a short discussion of the forms being used for exchanging Routing Policy and configurations. These forms are an initial attempt at a mechanism to detect certain classes of global policy inconsistencies. RIPE commented that the forms should be network oriented, and not AS oriented. Attendees Nagaraj Arunkumar nak@3com.com Tony Bates tony@ean-relay.ac.uk Jordan Becker becker@nis.ans.net David Bolen db3l@nis.ans.net Henry Clark henryc@oar.net Michael Craren mjc@proteon.com John Curran jcurran@bbn.com Osmund de Souza osmund.desouza@att.com Dino Farinacci dino@cisco.com Stefan Fassbender stf@easi.net Dennis Ferguson dennis@mrbill.canet.ca Peter Ford peter@lanl.gov Vince Fuller vaf@stanford.edu Der-Hwa Gan dhg@nsd.3com.com Elise Gerich epg@merit.edu Eugene Hastings hastings@a.psc.edu Kathleen Huber khuber@bbn.com Steven Hubert hubert@cac.washington.edu Paulina Knibbe knibbe@cisco.com Mark Knopper mak@merit.edu John Krawczyk jkrawczy@wellfleet.com Tony Li tli@cisco.com Daniel Long long@nic.near.net Matt Mathis mathis@a.psc.edu Dennis Morris morrisd@imo-uvax.disa.mil Brad Passwaters bjp@sura.net Robert Reschly reschly@brl.mil Erik Sherk sherk@sura.net Roxanne Streeter streeter@nsipo.arc.nasa.gov Marten Terpstra terpstra@ripe.net Carol Ward cward@westnet.net Chris Wheeler cwheeler@cac.washington.edu Linda Winkler lwinkler@anl.gov Jane Wojcik jwojcik@bbn.com Paul Zawada Zawada@ncsa.uiuc.edu 3