Agenda Bashing MadDogs Update Deferred by Meyer GLOP Draft Status - draft-ietf-mboned-glop-addressing-00.txt - 233/8 with AS in middle 2 octets - 233/8 assigned by IANA - runs out in June, can be renewed Update: transition from GLOP GLOP addressing in 233/8 WG last call Thaler: Why look at this for 232/8 - Is 64 bit addressing an issue? Meyer - possible that there may be single source models that don't use extended addresses. (not ESL model) Meyer - We should *not* move GLOP to BCP for now. Ron Roberts - can we determine which addresses are generating high bandwidth? Meyer - not addressed yet - submit a draft. Randy - conflict between last call and the experimental nature of address space Meyer - leave it experimental. Anycast RP draft status (DMM slides) - name change - foo-anycast-rp-00.txt within a domain, deploy more than one RP for the same group range give each RP the same IP address assignment - anycast address sources and receivers use closest RP sources known RP-RP via MSDP (convergence slide) . Lots of deployment (i.e. Sprint) . PIM-SM slide (Sprint) generic anycast approuch (level3 DNS servers) (unknown) ? about MSDP events of last week DMM - SA storms caused by SA looping - caused by ability to set defaults routes - instance of mis-configuration causing storm ... (blah) need tools (peer-RPF for upwards trace) Comments: Pusatari - Terminolgy a problem Pusatari: Section 6.2 interactions single-homed MSDP speaker always accept ...seems like a bad thing doesn't say which address to use - some practices active but not documented 6.2.4 default peers - what are default peers? seems like a bad idea 6.2.4 - bad because it shuts off split horizon Dino - can't do default route since you might not be directly connected to MSDP peer - Pusatari: could tunnel resolution - take to MSDP working group will import MSDP desicions back into this group MRM status: (Kevin Almeroth) - mrm manager - mrm agent o Test Sender (TS) o Test Receiver (TR) heartbeat - message types (TSR) (TRR) - TSs generate packets (DT - packet ? contol or data? - KA - data) - TRs are polled or generate alerts (loss thresh) - beacon messages refreach state o can also updaye (unreliably) state Operation: key is holdtome - tells sender how long to send - tells R how long to listen mechanism - TRs start test when hearing 1st packet or after timeout value specified in the TRR - holdtime of 0 stops test Goals: intra-domain management tool (NOC) - reachablity combined w/other tools inter-domain in the style of sdr-monitor - http://imj.ucsb.edu/sdr-monitor/global/ Status: - new draft - growing vendor support o cisco version in S code train o Nortel & Junier in the works - End host implamentation o imj.ucsb.edu/mrm/ - manager support in development (mmon) Changes: - name change "reachablity" - admin changes + clarification - add a length field (16 bytes - 2048 kbytes) - monitoring existing groups is put on hold - support for beacon messages is optional in the manager Open Issues: - best way to do security (?) - how to monitor existing groups - NACKs for TSRs/TRRs o needed for inter-domain or dynamic conditions - better support for alternate packet types - good justification in the face of SNMP Questions: Thaler: - Keep thinking about SNMP (answers 1st 3 bullets) - read RTP MIB (common goals) - recommend: framework independent of control protocol Fenner: RTP MIB has support for TR could evolve to supporting all MRM requirements (TS etc.) Fenner: Not sure what linkages - but will be backwards compatable Almeroth: DT what piece of draft do you want moved to an archetecture? Thaler: - Abstract out transport, security model, test data format and - TR/TS model. Examine the transport protocol in seperate doc (RTP MIB integration) Fenner: Agree with Thaler Connecting Multicast Domains (Brad Cain): draft-cain-mcast-connect-00.txt Intro: providers PIM-SM/MSDP?MBGP enterprises - PIM-DM/DVMRP - MOSPF - PIM-SM describes options - Border Routers cache architecture protocols submit (s,g) forwarding entries to shared table . several ways to bring transit sources into stub domains - domain wide reports - recievers are senders heuristic (*.*.RP) MSDP on boarder Flood-and Prune protocols Stub transit examples (see slides) PIM-SM to PIM-SM enterprises use as intra-domain routing protocol connection options are - standard MSDP MSDP to provider - stub domain may use providers RPs - stub domain becomes extension of provider PIM domain - may have its own RP for admin scoped addresses Comments Thaler: Unfortunately, admin scoped domain not allowed in PIM2. Cain: Filter RP data sets at the border Thaler: Need a draft in PIM WG - I'll help. MPSPF in stub easy to connect because - group LSAs provide global group membership info - >> MOSPF Summary: need comments on draft - dvmrp<-> MBGP - default route use in stub section on BiDir tress - BGMP and PIM-SM bidir publish as informational doc SDP and Multicat scoping (Colin Perkins) Problem SDP containd a mc address Solutions - announce session w/ SAP - include a scope indentfier How to choose the scope identifier MZAP has a scope name would require - manditory - name allocation MZAP zone ID - unique (may change w/time) - may be static enough best option now Concluding - problem that needs solving? - use zone ID - any alternatives Le MUR - Multicast relays in firewalls (Carsten Bormann) RFC2588 - don't let bad packets in - don't let confidential packets out Unicast considerations - need to be able to talk back to source a. multicast everything b. allow unicast path back - external sources - need dynamic hole punching address translation o little need to translate multicast addresses o translate unicast address Running firewall: Packet filters - can't unicast back to source - rewrite source address (NAT) - use HTTP for feedback (protocol change) Proxy firewalls - terminate in firewall - run internal instance of mc - translate between domains - need way for proxy to represent multiple internal members Multicast relays - RTP translator idea - relaying destroys sender addresses - congestion control info Unicast relay - NAT/PAT translate between service models