Presence and Instant Messaging (PRIM) BoF
Meeting -- IETF50 -- Monday, March 21, 2001, 15:30-17:30
Official note-taker: Thanos Diacakis
Chair: Florencio Mazzoldi
Co-chair: Hiroyasu Sugano
PRIM core features
There were no issues with the agenda.
Florencio presents status of PRIM. Revision of the protocol draft submitted, finalizing the charter and working on open issues.
Thanos Diacakis presents differentiators of PRIM from other protocols.
Simplicity, client-server-server-client architecture, atop of TCP, minimalist approach.
Asks group for hum whether this is on the right track.
Hum is a yes.
Comment: This is on the right track
Brief discussion about the differences with the SIP approach.
Comment: Why do we need SIP? We already have it for other things.
J Rosenberg: Already have SIP for Presence
Falstrom: This is not a SIP vs. PRIM discussion.
Comment: Go back to explaining/examining SIP
Comment: SIP is not a standard yet
Comment: Presence from different sources, therefore no single type of transport.
TCP is already used; we need it.
Mark Day: Cannot do hum until we get an idea of unresolved issues.
Comment: Like PRIM because it's lightweight.
Comment: PRIM & firewalls. How does that work?
Thanos Diacakis: Basically same as with any other of the proposals, you need to open the port.
Sugano goes through slides of open issues:
- CPIM alignment
- Multiple UA behavior
- Granularity of data updates
- Server-Server authentication
Requirement: align with CPIM as much as possible
Comment: Wrapper is XML, how does that work?
Comment: No, that's for presence
Comment: Fetch in CPIM is by subscribe with zero length. How does PRIM solve this.
Sugano: Can replace fetch with subscribe-zero.
Vassilis Polychronidis: What does "Comply with CPIM as much as possible mean?"
Mark Day: Examples of where compliance is not possible.
Florencio Mazzoldi: none at present
Thanos Diacakis: Don't expect to ever be non-compliant
Mark Day: PRIM is CPIM over TCP
Florencio Mazzoldi: PRIM is CPIM over TCP plus client-server
Comment: PRIM is centralized, not distributed approach
Thanos: No, PRIM is distributed
Mark Day: Point is moot as you can co-locate client and server
Comment: Need to trust other domains' servers
Vassilis Polychronidis: Can use encryption
Going through charter:
Replace "It will place well known..." with "Using existing architecture".
Jonathan Rosenberg: on "does not use SIP or BEEP", but uses DNS, etc. How do you position it?
Mark Day: How about take CPIM and add the least possible amount of things to make it work?
Consensus on Mark's comment.
Comment: Problem with end of second bullet.
Thanos: Add word "distribution" at the end of the last sentence.
Aoki: Support for group messaging?
Thanos Diacakis: Not in requirements.
Vassilis Polychronidis: How are milestones affected by CPIM?
Thanos: Add "subject to CPIM completion on relevant milestones".
Hum to continue with this effort.