Editor's Note: 11/23/92 CURRENT_MEETING_REPORT_ Reported by John Moy/Proteon Minutes of the Open Shortest Path First IGP Working Group (OSPF) The OSPF Working Group met at the November 1992 IETF in Washington, D.C. The meeting began with some administrative details. For those that had missed the announcement, it was mentioned that RFCs 1370/1371 had been published, officially making OSPF the recommended IGP for the TCP/IP Internet. Next, an informal poll was taken concerning attendance at next year's Amsterdam IETF. Based on that poll, the OSPF Working Group will probably not meet at that IETF. Lastly, there was some discussion of maintenance of the OSPF mailing list (trantor.umd.edu needs a forwarding record for ospf-request). The last call for comments was started on four documents. These documents have been stable for a number of months, and we hope to have them issued as RFCs before the next IETF (the latest versions of all of them had been issued as Internet-Drafts before the meeting): o The OSPF V2 specification. Two changes have been made since the last meeting. First, the definition of the Link State ID in type 3 and 5 LSAs has been relaxed to support CIDR (supernetting), allowing one or more of the Link State IDs host bits to be set. Since this will soon be an operational issue, vendors were encouraged to modify their implementations accordingly as soon as possible. It was decided to make the default route a special case, requiring both its mask and its Link State ID to be 0.0.0.0. The other change to the spec was a reminder not to set the ``AS boundary router'' bit in router-LSAs for stub areas. o The OSPF MIB. Changes had been made to allow host routes to be deleted, and to support the new NSSA and multicast routing options. At the meeting, it was decided to clear up the definition of the ospfAreaLsaCksumSum object, making it clear that it included optional LSAs (like the group-membership-LSA) and therefore was not necessarily the same for all routers in an area. Also, to better support supernetting, it was decided to include a network mask in the Link State Database and External Link State Database tables. o The NSSA option. At the meeting it was decided to change the document to allow summarization of type 7 LSAs into a single type 5 LSA at NSSA boundaries. o The OSPF Trap MIB. This document has remained unchanged since the last meeting. A number of issues were brought up by Robert Ching, on behalf of the OSPF Forum. Most were just requests for technical clarification (addressed in the meeting, but omitted from these notes in the interest of brevity). There was desire for an OSPF/RIP transition document (any volunteers?). Also, there was some confusion over the way OSPF 1 represented serial lines. As John Moy explained, they are represented in router-LSAs as a direct connection to a neighbor, with each neighboring router advertising the other's serial line address as a (stub) host route. This encourages pings to a serial line address to actually traverse the serial line. However, two other representations are also possible: each neighboring router advertising its own address as a host route, or each neighboring router advertising a stub route to a subnet that has been allocated to the serial line. It was pointed out that the latter representation had the problem that traffic addressed to a non- existent host on the serial line had a tendency to loop until its TTL expired. Osmund deSouza outlined a proposed usage document for OSPF over Frame relay. Requiring no protocol changes, this document would allow a Frame relay network to be configured as an arbitrary collection of NBMA networks, numbered and unnumbered serial lines. Tom Pusateri presented his document on running IP multicast over 802.5 networks. A functional address (03-00-00-20-00-00) has been allocated, and Tom's document mandates that the token-ring address be configurable as either the all-ones broadcast MAC address (current practice), the new functional address or a group address (for possible future definition when token ring controller support is available). This document should soon be published as an RFC. John Moy led a discussion of his proposal for how to deal with OSPF database overflow. It was decided to: 1. Exempt default routes from the limit calculation. 2. Automatically regenerate routes that have been earlier flushed due to database overflow (this regeneration will be done after some random interval between 1 minute and a configurable upper bound, with an option to completely disable the regeneration) and, 3. Set the LSA limit (which must be the same through all routers) through SNMP. Hopefully we will have a document describing this in detail next meeting. At the end of the meeting, mention was made of two possible new work items: 1. A scheme to use the OSPF tag field and a new LSA type to replace IBGP and, 2. A new authentication type using something like MD5. These are possible subjects for the next meeting. Attendees 2 Fred Baker fbaker@acc.com Ken Benstead kbenstead@coral.com Jeffrey Burgan jeff@nsipo.nasa.gov Dean Cheng dean@sun2.retix.com Robert Ching rching@nat.com Rob Coltun rcoltun@ni.umd.edu Osmund de Souza osmund.desouza@att.com Vince Fuller vaf@stanford.edu William Haggerty haggerty@ctron.com Jonathan Hsu brenda@penril.com Akira Kato kato@wide.sfc.keio.ac.jp David LeRoy dleroy@mitchell.cit.cornell.edu Tony Li tli@cisco.com Olli-Pekka Lintula olli-pekka.lintula@ntc.nokia.com Robin Littlefield rlittlef@wellfleet.com Kent Malave kent@bach.austin.ibm.com Jun Matsukata jm@eng.isas.ac.jp David Meyer meyer@oregon.uoregon.edu Douglas Miller dmm@telebit.com John Moy jmoy@proteon.com Julianne Myers jmyers@network.com Laura Pate pate@gateway.mitre.org Thomas Pusateri pusateri@cs.duke.edu Manoel Rodrigues manoel_rodrigues@att.com Paul Serice serice@cos.com Erik Sherk sherk@sura.net Roy Spitzer roy.spitzer@sprint.com John Tavs tavs@vnet.ibm.com Paul Traina pst@cisco.com Iain Wacey cat@pluto.dss.com James Watt james@newbridge.com Luanne Waul luanne@wwtc.timeplex.com Douglas Williams dougw@ralvmg.vnet.ibm.com Linda Winkler lwinkler@anl.gov 3