IETF78 Maastricht Drinks WG meeting =================================== notes by Brian Rosen 1. Doc status use cases/rqmts nearing WGLC, reviewers wanted, late on milestone, hopeful to request publication by September SSProv & transport- new WG items, milestones in November 2. Use cases & protocol requirements - Sumanth 2 versions since Anaheim JF - Peering Organization element associates a peer with a set of routes = selective peering David - Policy is orthogonal to peering and ingress points JF - Policy is not in the protocol, but routes dependent on peer is essential Jon Peterson - routes to administrative entities is fundamental to this work Syed - Trying to understand if peer getting multiple routes chooses David - Route group choice may be dependent on destination Sumanth - not saying you have to make a choice, but providing the option JF - both the originators and terminators have choices and terminator may see multiple routes Alex - there many be many routes, the mechanism described selects some, the destination selects one. No applying policy on behalf of someone else. The point here is to control visibility by peer. Daryl - Agree with Alex and JF. If I specify order and preference I expect dest to conform. Otmar - 2 questions, do we have any idea how often changes will be made? Is this for short term or long term traffic engineering Sumanth - real time changes are allowed. Don't know what users will do Otmar - the use case of name server response, assumes enum only and doesn't fit in with the rest of the doc Daryl - need both real time and static routing changes Brian - can make the NS response more general = refer to some other database David - where and why, provision the where, don't specify the why Otmar - so we don't want, for example, time of day routing Alex - correct, no phase of moon work yet Sumanth - could allow information to be passed without defining what it means JF - Implementations exist where policy drives off the shelf ENUM resolvers. So we let that happen downstream, don't do policy here Alex - allow policy to be transported (later) Otmar - do all names map Alex - does that means wild cards in identifiers Daryl - if you put in a URI, you do have all you need for DNS routing JF - You have user@domain, but domain may not resolve with 3263, may need a survey of ISPs Otmar - very uncommon to resolve on the Internet Sumanth - allow wildcards Otmar - remind me of the link between public identity and SED record, its a hack Ken - user specific SED records are sometimes needed JF - not service provider centric, allows a provider Nick - can the public ID be associated with both? Sumanth - yes Sohel Khan and Jon Peterson agreed to do a review of the draft before we go to WGLC. 3. SPPP - Ken Otmar - what is a dependent peer? Ken - terminating David - (re Carrier of Record Claim) - is there an announcement mechanism Ken - standard poll vs Pub/Sub issue JF - This is a requirement that you brought in. Mechanism can be improved. Date as optional at least Sumanth - Agree that notification would be good JF - push is not in scope for this version of protocol Sumanth - but we can document the need David - future agenda item JF - could we document a use case for this and clearly label for further study Otmar - on Carrier of Record Claim, ?notetaker didn't really understand the question, but the answer was the resolution protocol deals with it Syed Public Identity abstracted Otmar - why email Sumanth - can use other things, email is provided as an example Alex - not email portability mechanism JF - provision a TN and other meaningful identities, SMS, vCard are examples besides email Alex - sensitivities on email, perhaps use a different example JF we could delete it Jon - This is in the charter, right? Work on ENUM first (TN in, route out). That should be the main use case. Other things possible, but not now. Don't need to take use case out. Like "AbsoluteURI" Otmar - the Absolute URI is an output, not a query key, right? Syed - yes Abstraction of routes - NAPTR, NS and URI David - clarifying - things like SPIDs would be encoded as URIs Sumanth (referring to range) - need a prefix Brian - need to encode a prefix, not just a range JF - treat everything as a prefix Hadriel - you need to specify it as a prefix Sumanth - change to name in Add Egress Route Alex - comment on what happens next Syed - please review, more revisions coming Otmar - having trouble mapping to Sumar's presentation, route group id and route information in one step? Syed - Yes Sumanth - Need tighter connection (update to use case to match protocol) JF - comment on next steps. Heard comments that too much on design team calls, not enough on the list. We put things on the list, but don't get much feedback Alex - either the threads on the mailing list are too complex, or the people who care are on the design team Gonzalo - ml also defines progress, so continue Otmar - small working group, just move to ML JF - don't agree - we meet weekly, complete AIs, want to keep design team Keith Drage - must solve "publicity problem", must report regularly on what the design team is doing Sumanth - we do that Alex - always report minutes at least JF - we meet weekly, assign AIs, but cut some slack Alex - posting minutes is good, everyone has a day job 4. SPPP Transport - Ken 5. Thoughts from an Operator's Perspective - Nick Russell Hadriel - Lots of good ideas. Probably not for Drinks. Dispatch?? Maybe a BOF. Some of this looks very hard - getting part of a domain to route to one carrier another part to a different carrier JF - Some work in Speermint is relevant Jon - Agree, Drinks not appropriate, Speermint would have been better, but probably a BOF is a good idea. 6. Global SPID JF - we have a specific suggestion, enterprise numbers Hadriel - larger than drinks, may be a new scheme Sumanth - one way is a URN, Penn Pfautz was going to do a draft JF - Need to write some text. It's not done. Re Hadriels comment. But we need it in SPPP badly, can't wait a year. Use enterprise numbers now Nick - how does this relate to ITU-T Global SPID? Brian - use a name space identifier and a value Otmar - urn could be self typing JF - Like Brian's idea 7. AOB Otmar - what does demise of SPEERMINT mean to us Jon - Architecture doc will be finished, and we need that, everything else is fine