CURRENT_MEETING_REPORT_ Reported by John Moy/Proteon Minutes of the Open Shortest Path First IGP Working Group (OSPF) The OSPF Working Group met on Tuesday, March 30th at the Columbus, IETF. The Minutes of that meeting follow. The meeting began with a discussion of the six documents that the Working Group had in progress: 1. The latest draft of the OSPF V2 specification was reviewed. The only change that had been made since the last meeting was to handle the following case: A router becomes Designated Router, originates a network-LSA, then later restarts with a different Router ID and becomes Designated Router again. At this point, the router originates a network-LSA having the same Link State ID, but different Advertising Router, than the previous network-LSA. The fact that these two network- LSAs can both exist in the OSPF domain concurrently can confuse the Dijkstra calculation. Text has been added to the specification ensuring that the network-LSA originated before the last router restart will be flushed. After some discussion, it was decided to submit the latest draft (which had been published as an Internet-Draft) to replace RFC1247 at the current standards status (Draft Standard). 2. The OSPF Trap MIB had remained unchanged, except for editorial comments, since the last meeting. It was decided to submit this to be published as a Proposed Standard RFC. 3. The OSPF NSSA area specification was reviewed. A problem was found in the aggregation of multiple type 7 LSAs into a single type 5 LSA, involving the choice of metric. Two solutions were discussed: o Have the type 7 LSAs always take precedence, and o Set the metric to be the largest of any of the component metrics. Rob Coltun will investigate these options further. 4. Osmund deSouza presented a document describing how to run OSPF over Frame Relay. The document describes how to split the Frame relay PVCs into collections of OSPF Point-to-point networks and NBMA networks. A comment was made that treating PVCs as unnumbered links was problematic, due to the inability to assign an ifIndex to individual PVCs. It was decided that after adding comments, the document will be submitted for publication as an Informational RFC. 5. No progress had been made on the OSPF MIB, which needs some additions before it can be republished. 6. No progress had been made on the ``OSPF Database Over-flow'' 1 document. The ``OSPF for SIP'' Internet-Draft written by Christian Huitema was summarized as: regular OSPF, running over IPv4, with two additional LSAs to import SIP information and an additional bit in the router-LSA to indicate SIP capability. This was intended to allow a more or less seamless migration from IPv4 to SIP, after which a native OSPF for SIP would be defined. Detailed discussion of the Draft was carried on in the SIP Working Group. In fact, it was decided that all detailed discussions of OSPF in IPv7 would be carried on in the appropriate IPv7 working groups. Dennis Ferguson then presented an overview of his ``OSPF external attributes'' proposal, which is an addition to the OSPF<->BGP routing interchange and can be used as a substitute for Internal BGP. Dennis also presented operational statistics from the NSFnet that indicated his proposal would be quite efficient. Discussion indicated that efficiency, measured in terms of the percentage of the database dedicated to this scheme, would decrease when CIDR was deployed. Tony Li mentioned that it will also be necessary to know whether all routers participating in the ``OSPF external attributes'' are BGP-4 speakers, or whether some are BGP-3 speakers, in order to decide whether BGP aggregation should be done. Most of the discussion then centered on the problem that, since the external attributes (type 8 LSAs) can only be flooded through supporting OSPF routers, it is possible that the database of external attributes could get out of synch with the type-5 LSAs (which in turn could lead to problems in BGP routing). Dennis suggested three ways of dealing with this: 1. Choose the Link State IDs for type 8 LSAs in a random fashion, so that lack of synchronization would be obvious. 2. Potentially run Dijkstra a second time to ensure that there is type-8 flooding connectivity between BGP speakers, or 3. Change the document so that ``most'' routers must be capable of flooding type 8 LSAs. No decision was made on these options. Finally, it was noted that a combination of Internal BGP and the new ``OSPF external attributes'' cannot be run since the tag field in the OSPF type 5 LSAs would then have two conflicting requirements. Tom Pusateri presented an outline of a proposed RIP to OSPF transition document, based on a talk he gave at INTEROP. He solicited suggestions for additional items to cover (send to pusateri@cs.duke.edu). Suggestions given at the meeting were: 1. Warning against running OSPF and RIP in parallel. 2. Give an example of a real, non-trivial network and how to 2 transition it. 3. How to do address assignment. 4. How to decide what's in an area. Lastly, a conflict between OSPF and Router Requirements was mentioned. Router Requirements states that you can't follow the default route to get to subnets. This rule doesn't work for OSPF stub areas, and several people mentioned that the rule, while consistent with RIP, shouldn't really apply to other protocols. A different rule, along the lines of: ``when aggregating, create a discard route for the aggregate'', was suggested. Philip Almquist, the editor of the Router Requirements documents, was present and participated in the discussion. Attendees Philip Almquist almquist@jessica.stanford.edu Dennis Baker dbaker@wellfleet.com Fred Baker fbaker@acc.com Jim Beers Jim.Beers@cornell.edu Nutan Behki Nutan_Behki@qmail.newbridge.com Richard Bjers rich.bjers@uc.edu John Boatright bryan_boatright@ksc.nasa.gov Robert Calderon calderon@noc.ans.net Douglas Carson carson@utcc.utoronto.ca James Cassell jcassell@dsac.dla.mil Rob Coltun rcoltun@ni.umd.edu David Conrad davidc@iij.ad.jp Wayne Cullen wnc@netlink.com Kurt Dobbins kurtdob@ctron.com Kishan Dudkikar kishan@icm1.icp.net Dennis Ferguson dennis@ans.net Paul Franchois paulf@bldrdoc.gov Christine Fredenburg cfredenburg@dsac.dla.mil Vince Fuller vaf@stanford.edu Darren Griffiths dag@ossi.com Patrick Hanel hanel@yoyodyne.trs.ntc.nokia.com Jeffrey Honig Jeffrey_C_Honig@Cornell.edu David Jacobson dnjake@vnet.ibm.com Zbigniew Kielczewski zbig@eicon.qc.ca John Krawczyk jkrawczy@wellfleet.com Duane Kuang duanek@kalpana.com Tony Li tli@cisco.com Robin Littlefield rlittlef@wellfleet.com Glenn Mackintosh glenn@canet.ca Jamshid Mahdavi Mahdavi@a.psi.edu Glenn Mansfield glenn@aic.co.jp Jun Matsukata jm@eng.isas.ac.jp James Miner jjm@fibercom.com John Moy jmoy@proteon.com Julie Myers jmyers@network.com Shannon Nix sdn@netlink.com Zbigniew Opalka zopalka@agile.com 3 Ayal Opher aopher@synoptics.com Joe Pagan jrp@afterlife.ncsc.mil Thomas Pusateri pusateri@cs.duke.edu Edward Reed eer@cinops.xerox.com Ben Robinson ben_robinson@vnet.ibm.com Benny Rodrig 4373580@mcimail.com Manoel Rodrigues manoel_rodrigues@att.com Hal Sandick sandick@vnet.ibm.com Shiva Sawant shiva@synoptics.com Kanan Shah kshag@cmf.nrl.navy.mil Andrew Smith asmith@synoptics.com Martha Steenstrup msteenst@bbn.com Steve Suzuki suzu@fet.com John Tavs tavs@vnet.ibm.com Marek Tomaszewski marek@net.com Kannan Varadhan kannan@oar.net Linda Winkler lwinkler@anl.gov Jane Wojcik jwojcik@bbn.com 4