Minutes, SIPPING WG IETF 55 Edited by Dean Willis Meeting Notes by Vijay Gurbani Meetings in Atlanta, GA, USA Two Sessions: Wednesday, November 20 at 1530-1730 Thursday, November 21 at 1300-1500 CHAIRS: G. Camarillo Dean Willis R. Mahy Wednesday, November 20 at 1530-1730: Salon III =================================== Agenda Bash -------------- No issues. Status and Charter, Chairs --------------------------- New co-chair Gonzalo Camarillo introduced. Status report given by Rohan Mahy. Discussion: Scope and home of filtering work. Conclusion to pursue rate limiting in SIPPING as a generic topic and that individual event packages should specify their own filtering requirments. Group agreed to consider a framework draft for rate limiting, and Aki Niemi volunteered to attempt a first draft. Registration Event Package Open Issues, Jonathan Rosenberg --------------------------------------------------------------- Statuess: In IETF LC. One change: add shortened event to the state machine. Issue: default expiration matching default registration expiration (default lifetime: 3600 sec, same as default registration time). Made people nervous due to race conditions/synchronization problem. Proposal: change to 4200s. No opposition shown Issue: XML vs sipfrag: why convert reg msg to XML? Discussion: Extensive on pros and cons and costs. Poll: XML vs anythying else. Consensus: XML. SIP-T Related Issues, Jon Peterson ------------------------------------ Status Current work done; SIP-ISUP is in authors 48 right now. Overlap I-D is still in progress; not a general solution -- there is none. Issue: Overlap:.What should we do? SG11 is considering using this from the IETF. Discussion: Wide ranging discussion of requirement and rationale for overlap and whether the overal draft should ve published and if so on what track. OPtion of letting ITU deal with it raised. Noted that if done in IETF, we will have serious security concerns. Conclusion: There doesn't seem to have been any conclusion. Topic: New directions for SIP-T: Issue: SDP mapping: draft by Gonzalo on early media and other SDP mapping aspects: code mapping and absence of SDP (in case of 3PCC). Q.931->SIP mapping: Discussion: widely implemented without a standard, not sure what value a standard will add to it. Issue: Privacy-identity mapping for SIP-T: Discussion: SG11 already looking at this. We are looking at the P-Asserted-Identity version, not the long- term identity we discussed in sip yesterday. Suggestion that we at least do the privacy mapping for long term identity and that this could be informational. Suggestion that we make this an explicit non-goal and that we minimize "educational" I-Ds. Conclusion; Jon Peterson will attemp Privacy-mapping I-D. Group will not do any Q.931 work until at least next meeting. SDP mapping will go to list for discussion. Poll: Consensus: do people agree on working on document on how to use existing tools on providing early media in SIP? A document that explains how to use UPDATE and other tools to do early media?. Result: consensus to do this work. Poll: Should we do 3PCC/SDP interworking with PSTN - No one hummed. Discussion inidicated that there is no clear understanding of the problem and we need to get a individual I-F writeup. Message Waiting Package Open Issues, Rohan Mahy -------------------------------------------------------- Changes since last version reviewed. Discussion: Suggested that a couple of things are complex enough to be non-trivial and not important enough to be useful. Given that we do any other events stuff shall we align it with event style description rather than having a restrictive count only description? Poll: Continute with text body: No objection. Concluion: Document will be revised to account for new caller-prefs. Transfer and Service Flows Open Issues, Alan Johnston ---------------------------------------------------------- Topic: Transfer Changes since last version reviewed with slides. Issue: A Refer-To URI needs to allow the transferee to reach the transfer target. One way is to put Route header in a Contact, but this is illegal as per rfc3261. Discussion: This problem keeps popping up in many places; in conferencing it happens as well. Better not to solve it here, but rather do it in a separate I-D. Poll: Any voluntters? None. Remaining work: close open issues, track REFER as it completes, add uses showing Referred-By. Request History Open Issues Mary Barnes ------------------------------------------- Slides presented are available in proceedings. Discussion: It might be possible to use this in conjunction with Jiri Kuthan's SIP diagnostics. Maybe we can use this work to apply in that context. Issue: ABNF issues: explicit counter: plain counters do not work due to forking options: indexiing using a dot delimiter to reflect the depth of forking Forking: need to add more scenarios Privacy: if you need privacy, use draft-ietf-sip-privacy-general Security: primary concern is in regard to a rogue app/proxy challenging HI entries. Discussion: You are already assuming some level of trust in these intermediaries; i.e. they are proxying the request in the first place. It strikes me as a "sips" kind of thing; then you can guarantee that no rogue element has put anything in since all are trusted. Proposal: Maybe you can use the Auth-ID body draft. Idea is not to assume a weird trust relationship b/w all intermediaries. What you get is some optional information in a signed auth-ID body that asserts that proxy A retargetted to location Y. Maybe we should allow proxies to add bodies (they add headers). Comment : This is hard to spec right. Maybe we can start with writing reqs so we can spec it right. Poll; Ready for WGLC? Resolved to take to list, not enough people had read Poll: Do we feel that the reqs have sufficiently been presented in public that we are ready to ask SIP to consider a mechanism? Even split between waiting before implementing a solution and asking SIP to proceed. Diagnostics -- Jiri Kuthan ------------------------- Status: Narrowed the scope due to the discussion of mailing list; scope now is diagnostics of SIP routing only. Proposal: Let the UAS be talkative and disclose what they know about circumstances under which a request failed. Proxy Hints: disclose aggregated processing logic; mirror hints upstream so that troubleshooters located upstream can see it Issues: privacy concern. Proposal is to use sipfrag and include the diagnostic in sipfrag. Discussion: Might be better to have a standardized way of doing this to make it easier to implement and deal with. However, we have the same problem with email processing; the email server sends error messages that say where the message died. Proposal: Reqs developed for history would apply to this quite well. We should start with that and see how far it gets us. Conclusion: We should start working on requirments.. Thursday, November 21 at 1300-1500: Salon III ================================= Evaluation of IEPREP requirements on SIP, Chairs, Henning Schulzrinne ------------------------------------------------------------------------- Changes since last rev discussed. Issue: What should the process be? Does this draft cearly define requirements for extensions to SIP that can be taken to SIP WG? Conclusion: We will scheduel a WG review ASAP. We will not crosspost to IEPREP, but will ask a volunteer (Kimberly?) to take any discussion needed to IEPREP Status report on Conferencing design team, Alan Johntson -------------------------------------------------------- Status New smaller design team formed to try and get back on schedule and docs agree with each other. Each team member now responsible for one or more docs. First out will bea revised framework document that will include terminology. There are drafts of several documents, agreements on terminology and scope, and progress seems good. Poll: Anyone have problems with this approach? Suggestion that docs be out a month to six weeke before next IETF. Status report and open issues from Hearing Impairment design team, Gonzalo Camarillo ---------------------------------------------------------------------------------------- Slides presented are available in proceedings. Goals: Going beyond the reqs for the deaf; we are working on invocation of services that involve media manipulations including codec and language transcoding. Want to be opes aware. Deliveries: transcoding services invocation in SIP and the source and sink attributes for SIP (2 I-Ds). Possible future deliveries: labeling of transcoded media streams, caller prefs extensions. We have to think about these two before finalizing them. Issues: invocation in the B2BUA model as opposed to 3pcc model which is seen as quite complete. We use Route header field as if the B2BUA was a proxy. We did not feel comfortable with this approach. Maybe we can use something else like message encapsulation or REFER. Current decision is REFER: more opes friendly and has security built in. Disadvantage: more signaling. But generality wins. Discussion: the work may not have taken into consideration other requirements about transcoding. Consenus that these can be added as they are brough up on list. Poll: is plumbing draft (draft-camarillo...sink-00) needed? Discussion that it might overlap with conferencing and be able to reuse conference work or be input to conferencing work. Conclusions: co-ordinate with conf DT on overlap and ask SIPPING WG to advise on further path. App Interaction Design TeamT, Jonathan Rosenberg ------------------------------------------------ Status: Small team -- 4 members. What are we doing? Stimulus with SIP -- the way in which users interact with apps using media, markup and dtmf. Team is making good progress on understanding the nature of dtmf in these networks, but there is a general problem here: event reporting. Produced 3 documents (framework, KPML, and interaction) and inherited one on interaction requirements. Issue: infamous SIP vs HTTP debate -- who carries the information? Document clarifies difference between rfc2833 and using SIP or HTTP to carry these. It deliniates the UI from the app. Comment: If you have markups that assume HTTP they will use it; if you have markups that assume SIP, then they will use SIP. The language construction is different. Issue: Focus determination for KPML: when a user enters a digit, which KPML amongst the ones from various apps does it apply to. Current draft says send it everywhere -- same as PSTN does. Architectural solution: have a central controller model which runs a super app and make decisions for the user. Is this adequate? Any other ideas? No discussion noted. Issue: Error reporting for App-Info: what if App-Info URI is invalid? Options: no error reporting, only send App-Info in requests (where a error response can be generated), error reporting URI (infinite recursion problem). No discussion noted. Issue: indicating script termination -- should we or not? Discussion: This is complicated, and overlaps with IMAP issues Comment: A lot of it is problem statement: we want to do what HTTP does, but get started in some other way. Issue: Barge-in ==> framework has feature that allows IVR barging to be pushed to endpoint. Problem is how to detect when its OK to start playing media again? Need some media synch. Markup bits don't work; PT changes may work; others? Further discssion referred to list. Discussion about potential new work in QSIG and Q.931 interworking, Jon Peterson -------------------------------------------------------------------- Discussion Some interest in going forward with this in Yokohama IETF. Not a lot of discussion since then on list. As per yesterday's SIPPING WG, we would also prefer to do some SIP->Q.SIG mapping. John Elwell has released a new version of I-D since Yokohama; changes: use of P-Asserted-Identity in mapping to CLIP/CLIR and strengthened security text following the SIP-ISUP draft. Not a SIPPING WG item yet, any comments? Keith: Minor nit: it is QSIG, not Q.SIG. Poll: Any objections to make this a WG item? No objection noted. FYI and Discussion on iCal usage in SIP, Pekka Pessi ---------------------------------------------------- Slides presented are available in proceedings. Status: iCalendar is a calendaring information exchange format based on vCalendar. iTIP is an abstract protocol for interaction between calendar UAs. It has methods like PUBLISH, REQUEST, REPLY, CANCEL, ADD, etc. iCalendar use with SIP: off-line invitations, describing SIP tele-conferences, indicating schedule and timezone, notifications for schedule, conference, etc. Procedure :Map iTIP to SIP ==> iSIP. iSIP is very similar to iTIP. Open issues: does iSIP fit in calsch? It does not change how SIP is used. Submit new draft using standard iCalendar format? Specify iSIP apps separately: iCalendar for on-line conf? Discussion: Why do we have to carry this in SIP? Answer: SIP brings SIP addressing Discussion: One model is that this is some MIME type and I don't care. Then why do we have to define a mapping if we are carrying an object? Answer: It is an object and you don't have to do anything in SIP. However, some protocols provide optimizations for carrying the object. Do we want to do that here? Further discussion referred to list due to end of allotted time. Meeting adjourned.