BGMP met once at the DC IETF. These minutes were prepared by Bill Fenner. Dave Thaler spoke first, about debugging and monitoring of BGMP. First, a BGMP MIB is needed. Second, tools to debug the multicast RIB (S-routes and G-routes) are needed; a BGP4+ MIB with a PathAttr table per SAFI was suggested. On multicast traceroute: multicast traceroute already has rules for handling shared-tree protocols, so the only problem arises when crossing the inside of a domain may be problematic. This is handled by having the BGMP router encapsulate the trace across the domain to the upstream BGMP router when the MIGP could be problematic. Modifying "mrinfo" to include BGMP peering information was suggested, but it was agreed that "mrinfo" should be replaced by SNMP and so should not be updated with new functionality. Dave's proposals on monitoring action items: - add mtrace code for G-trace encapsulation - write BGMP MIB - Work on tools for debugging routes in M-RIB o unicast: topology databases, radb, etc o multicast: peering databases like mwatch o "find ANY legal path" via databases, then traverse that path to get an idea what is going on. Brad Cain talked about incremental deployment of BGMP and transition from PIM-SM+MSDP. BGMP sources can easily be injected into MSDP; the question is how to use a PIM-SM+MSDP domain as transit for BGMP. Such a domain is effectively multiple source-based-tree domains; one option is to use domain-wide reports to make membership known to the borders (DWR's would only be required at the RP's). Isidor Kouvelas talked about a problem with source-specific branches in BGMP; generally, the problem arises when a failure occurs on a sender-only branch after a source-specific tree has been set up. The BGMP authors are considering solutions to the problem.