NSIS IETF 56 MONDAY, March 17, 2003: Morning Session TUESDAY, March 18, 2003: Afternoon Sessions I
MONDAY, March 17, Agenda bashing - chair - 5 minutes Interim Meeting review - chair - 15 minutes RSVP discussion – Melinda Shore – 30 minutes Problems to solve – Henning Schultzrinne - 30 minutes NSIS Analysis doc - Jukka Manner - 15 minutes An Evaluation on RSVP Transport Mechanism - Ping Pan - 15 minutes
Interim Meeting Update - Consensus Seperate transport layer protocol state from NTLP state Soft-state refresh supported independent of transport protocol state. Extensibility should be provided (see something which a node does not understand, session should not fail) Seperate flow identifier and session identifier. Session identifier should be unique. Don’t overload ids Link-based congestion control is not a good thing. Scoping of requests should be avoided (see IPv6 mailing list)
Interim Meeting Update - Problematic Issues Uniqueness of session id Supporting multiple transport protocol options create complexity Security properties & needs Mobility support
TUESDAY, March 18, Agenda bashing - chair - 5 minutes NSIS Requirements doc - Marcus Brunner - 15 minutes NSIS Framework document - Robert Hancock - 20 minutes RSVP Security Properties - Hannes Tschofenig - 10 minutes NSIS Security Threats - Hannes Tschofenig - 10 minutes
Starting NTLP Work Melinda Shore & Henning Schulzrinne to co-author; Bob Braden & Michael Thomas to act as expert reviewers. Starting point RSVP ( ) Separate service specific stuff out Separate flow spec & session id Keep semantics of PATH (discovery & transport combined) Soft state Support of RSVP-proxies Open issues (testing consensus) Logically separate discovery & transport, in practice they could be piggybacked. Message fragmentation concerns Reliability concerns Optionally supporting other transport protocols Reduction of multicast complexity 2752 / policy issues Routing interaction (handling mid-session path splitting) Support for explicitly stating message directionality & hop-by-hop vs. end-to-end.