Editor's note: These minutes have not been edited. mboned 96.12.09 15:30 Dave Meyer Administrivia o mailing list mboned@ns.uoregon.edu (majordomo) Status o have some active drafts - pruning - administrative scope - will have new issue of seneria - Eliz Gentile's Deployment draft - RPSL draft on tunnels o also need to think about topology data base o need to discuss 500k limit and how to raise it o need to discuss dynamic address allocation Pruning Draft o was a communication problem between the author and the IESG. has been cleaned up and will go to BCP post haste. o it was a nice goal statement, but there is no enforcement. what could be done? state that the policy will be enforced. the reason this draft came about because people wanted to enforce but had no document. this provides that. Scoping Draft o was a de facto standard already o draft-ietf-mboned-admin-ip-space-00.txt o IANA has been asked to enshrine the number assignment o what kind of additional admin structure might there be? - flat has the advantrage that others may enhance structure - reserved scope addresses? SAP us to use highest multicast ddress within the scoped range. - but the address is large, so making some structure now would give us some hierarchical structure for the future - as border is bidirectional, hierarchy is out. - how about site, country, piece of wire, ...? i.e. geograpic - 239-255 was suggested as a local, probably site, scope o what about sparse mode or admin join protocols, how are these effected by administrative scoping? - explicit join model handles this - except boundary enforces scope - privacy can encrypt - except participant list - have to make sure that your RP is in same group as scope you are using. - why is admin scope mechanism linked to the particular protocol? - should we define a protocol which lets us discover the scope? Heartbeat Draft o SNTP as a Multicast Heartbeat - provide operational visibility for large auditoria on a dialup network - end users can subscribe to a stream o will not resolve - whether a particular dialup link is capable of multicast - deciding if a dialup line has connectivity tot eh mbone - solve this problem on the mbone, targetting a scoped environment o constraints - must work with IGMPv1 - join latencies have a high variation - membership state between application and stack may be out of sync - must not reqire configuration by end user o how it works - establish heartbeat which provides end-to-end control from media server to end client - quick convergence by having dialup servers, last hop before users, subscribe to group - put an SNTP packet into the message so you can use existing code, sequence numbers, rough time estimate o inplementation - client listner, continuous or only at startup - SNTP server, standard v3 with multicast capability - multicast devices o what they have done - ascend max 4000 . multicast address = G . source address S, mask = M . slot time = T . shift register length = N . required messages ove N slots = K . no authentication yet . if less than K messages in last N slots, SNMP trap generated . trap contains sysUpTome, G, T, N, K o trap backs off until it is down to 1/hr o benefits - a NOC gets a realtime picture - clients can decide to go uni or multi-cast - customer support has an extremely easy tool to debug problems while on phone o much discussion of necessity, benefits, join latency, ... o much discussion of why this presentation, BCP? protocol spec? o envisaged as a very short term solution. though it points out the value of being able to have an API which detects IGMP membership o they want to submit it as BCP o for a host which wonders, why not just send a multicast ping? RPSL Tunnels Draft o see RPS minutes (get from Sue Hares, who did it in detail) o there were actually good suggestions/corrections by folk who actually understood the semantics of multi-cast tunnels Discussion of o much discussion of benefits and problems of default rate limit o 3.9 will not have the default rate limit, but it will be required to specify something o MCI, BBN, ... do not rate limit o we can not include mtrace to add more info about routers, info is not necessarily known o should be a BCP making clear there is no global rate limit 12/10 (Scribe: Bill Fenner) Dynamic Address Allocation - Van suggests adding *.255/16 as the local scope range. - "No Rate Limiting" BCP will be written by Doug Junkins (NWnet). Agenda Topology Database Data sources? Van: mtrace listeners, but clients need to specify that response must be multicast back. Need a document describing this. Dave Thaler: Database would be useful for (i). Tracking loss over tim (ii). Generating network design (iii). Simulations (ave node degree, etc). Elizabeth Gentile Spoke on her draft. Math Crawford will be co-author.