IETF66 XCON WG Notes - Brian Rosen ---------------------------------------------------------------------------- Agenda Bash BFCP will be RFC4582 in AUTH48 University of Naples and Ericsson have interoperable implementations of BFCP. Univerity of Bremen also working on it. Way forward: Finish Data model, identify operations, then define protocol. ---------------------------------------------------------------------------- Editing Team proposal: Eliminate templates, put controls in appropriate places in the data model, allow extensions a la PIDF to describe additional controls Oscar sent a revised doc out Henning and Brian hold discussion on whether controls have semantic content beyond the "type"; is this a "user interface" description or a functional control definition. Chairs ask if we could accept the data model example presented as being reasonable without settling this. Henning believes that we must settle it. There are two irreconcialable models, so allow both. One control oriented, one semantically oriented. Roni Even says the data model wouldn't be different if its semantic or syntactic. Rohan Mahy says there is a piece of state (say a boolean for mute), may or may not be a control. State is represented in the data model. Wants Henning to show a counter proposal. Henning says fundamental difference is to define a new object you have to change the schema. Participants: Agreed that if a new control ("foo") was added, the SCHEMA (or Relax NG) would change. Cullen wants to know if an extension could concievably be handled by a client not directly implementing (coded to) the schema. Rohan says, there are some syntactic limitations, but to first order, we should be able to do that. Brian says that there are three states: not rendered, grayed or black/white corresponding to not implemented, policy doesn't allow it, or allowed. Need all three states distinguished. Adam says, the control could just not be present in the object to be "not implemented". Brian agrees that works ********************************************************************* ********************************************************************* *** *** *** Chairs ask for show of hands for agreement on this as basic *** *** direction: 20-30 hands for, zero against. *** *** *** ********************************************************************* ********************************************************************* ---------------------------------------------------------------------------- Gonzalo presents BFCP Connection Establishment Issues Connection Establishment Use offer/answer vis 4583, TLS Authentication BFCP establishment with Digest (no client cert) Provisioning Will be included in data model DNS Procedures Get a DN, but is it an FQDN, do we use SRV, A, AAAA Rohan asks if there is a well known port. Kinda likes SRV. Gonzalo says no WKP. Brian asks how the DN is learned = is it derived. If not, then A/AAAA is enough Adam says load balancing better with SRV Miguel says start simple A/AAAA Gonzalo has slight preference for SRV also ********************************************************************* ********************************************************************* *** *** *** Chairs ask if any objects to leaving out SRV for now; no *** *** objections. Consensus is to do A/AAAA only *** *** *** ********************************************************************* ********************************************************************* Connection Reestablishments Do we need a server mechanism Answer: no, client only Digest Usage Should we use Digest for first message only? BFCP is bandwidth efficient, don't want lots of challenges Roni asks how we do ICE on connection re-establishment Answer is use offer/answer Suggestion to use a count in the hash to skip the challenge Gonzalo wants to skip the whole signature ---------------------------------------------------------------------------- Mary Barnes presents Framework Updates: Fixed Dave Morgan nits, removed references to "user inteface abstractions", expanded examples, added sections on DTMF and Observing a conference. Work to complete: protocol issues, whisper, detail in sec 9, URI schemas, remove reference to templates Way Forward Send comments, trying to finish Adam discusses Relax NG rather than schema Henning says within RAI we should settle this one way (at least generally) Rohan - in favor of Relax. Lisa Dessault says, yeah, use Relax NG if it helps. Mary says she talked to her developer who liked the idea Miguel says his developers want to move to Relax NG = easier to understand, easier to design with Srivasta says need to make sure extensions to conference event package can be done easily when extending Relax NG ********************************************************************* ********************************************************************* *** *** *** Consensus to use Relax NG *** *** *** ********************************************************************* ********************************************************************* ---------------------------------------------------------------------------- Dial In/Dial Out/Refer Lists Presently 3 separate sections, proposed to consolidate one list, with attributes Brian says, okay, but does text say what happens when you remove someone from the list. Mary to verify that Framework is clear on this point Jonathan asks what happens if you have two attributes Adam says, appears twice in list ---------------------------------------------------------------------------- Meeting ends early Steve Noyes feels cheated, didn't get enough entertainment value from the meeting