Re: [martini] Proposed work split: tackle telephone-numberregistration first

[Paul Kyzivat <[e-mail]>]

(I sent this out with an incomplete thought I meant to delete. Here is 
revised version.)

Spencer Dawkins wrote:
> OK, this discussion continues to be lovely. Thanks again for listening 
> to the people you're talking to ("plays well with others" is 
> under-appreciated, once we get out of primary school, but *I* appreciate 
> it!)
> 
> After reading this thread down through Eric's note below, I have a few 
> requests, as co-chair.
> 
> We're talking about E.164 numbers like we can easily identify them. My
> understanding from Tuesday's call was that people on the call thought the
> only way we could unambiguously know that we had an E.164 number was if we
> had a tel: URI. If people on this list want to challenge that assertion, 
> this would
> be a great time to say so - please push back Real Soon Now.

I don't know that it is the *only* way, but it is an unambiguous way.

Of course sip URIs *can* contain E.164 numbers. The trick is in ensuring
that everybody concerned agrees that's what they are.

> If that's true, we're talking about using Tel: URIs. Dean has asserted 
> that this would require an update to 3261. I'm not a genius of 3261, but 
> some of you guys are - could you point us to what you think needs to be 
> updated?

That would come up *if* the intent is to *explicitly* register a TEL
URI. The key issue is from section 10.2:

       To: The To header field contains the address of record whose
            registration is to be created, queried, or modified.  The To
            header field and the Request-URI field typically differ, as
            the former contains a user name.  This address-of-record MUST
            be a SIP URI or SIPS URI.

But if the goal is to register using a SIP URI, and cause the "implicit
registration" of a number of TEL URIs, its not so clear that we have the
same problems with 3261.

The more important issue is "constructing" an R-URI from a registered
contact address as a part of implicit registration, and ensuring that
the resulting URIs are unambiguous at the pbx. As long as the implicitly
registered AORs are all TEL URIs, or sip URIs with a TEL user part, then
the pbx can know what to expect as R-URIs for incoming requests. And it
can know that the URI for a given telephone number will be the same
regardless of what SSP it uses.

(To ensure that, we would want to specify that the E.164 number was
represented in its canonical form, without any of the optional separator
characters.)

	Thanks,
	Paul

> My quick read turned up this text in Section 4, page 17:
> 
>   Registration is another common operation in SIP.  Registration is one
>   way that the biloxi.com server can learn the current location of Bob.
>   Upon initialization, and at periodic intervals, Bob's SIP phone sends
>   REGISTER messages to a server in the biloxi.com domain known as a SIP
>   registrar.  The REGISTER messages associate Bob's SIP or SIPS URI
>   (sip:[e-mail]) with the machine into which he is currently
>   logged (conveyed as a SIP or SIPS URI in the Contact header field).
> 
> Dean, Is this what you were thinking of? Are there other places that 
> also need to change?
> 
> As I read http://www.ietf.org/id/draft-peterson-rai-rfc3427bis-04.txt, 
> Section 2.1, this change would have to go through SIPCORE, because it's 
> an update to 3261. If you know of obvious land mines that we would be 
> walking into if we propose this change, please say so Real Soon Now. 
> (One obvious land mine is that DISPATCH isn't meeting in Anaheim, so the 
> proposal would have to be DISPATCHed on the mailing list in order to do 
> something before IETF 78 - are there OTHERS?)
> 
> Finally, if you believe that embracing tel: URIs is the wrong thing to 
> do, please say so, Real Soon Now.
> 
> The milestones that we committed to in 
> http://tools.ietf.org/wg/martini/charters have us adopting a working 
> group SOLUTION draft in January. We're not late yet, but it's the 28th 
> of January in my time zone. If we are late, where late is measured in 
> days, I can live with that, but if we are late, where late is measured 
> in IETF meeting cycles, that's going to be a problem.
> 
> Thanks,
> 
> Spencer, as co-chair
> 
> ----- Original Message ----- From: "Eric Burger" 
> <[e-mail]>
> To: "Dean Willis" <[e-mail]>
> Cc: <[e-mail]>
> Sent: Thursday, January 28, 2010 12:32 PM
> Subject: Re: [martini] Proposed work split: tackle
> telephone-numberregistration first
> 
> 
>> If you ask me (why can't I stay in retirement?), the E.164 situation  is
>> not only a simplification, but just about the only thing that is  
>> sensible
>> for MARTINI to work on.
>>
>> The E.164 situation is a clear, present, industry-is-begging-for-a-
>> solution problem.  It is so pressing, the industry will tell us (the
>> IETF) to get stuffed and do something stupid if we cannot deliver
>> something that is not at its core fundamentally broken.
>>
>> I do not think anyone can have an issue with mechanisms that translate  a
>> string of digits into an AOR that might cross domain boundaries.   
>> This is
>> the E.164 problem statement.
>>
>> If think everyone (with a brain) SHOULD have issues with a mechanism  
>> that
>> magically mungs "sip:[e-mail]" into 
>> "sip:[e-mail] "
>> while at the same time www.neustar.biz is really a neustar.biz  domain.
>> That really would be creating a parallel DNS, which, as one  of the major
>> DNS operators, would really, really be "less than ideal"  and have a 
>> whole
>> host of security and integrity issues.
>>
>> THEREFORE, I would propose that we split this work into two phases:
>>   Phase 1: Deliver an E.164 solution, preferably in the next few  months,
>> so the industry does not ignore the IETF.
>>
>>   Phase 2: Deliver the generic solution in the normal, 3-5 year IETF
>> cycle.  This generic solution may or may not encompass E.164 
>> numbers.   As
>> Doug Sauder points out, E.164 numbers and their infrastructure are
>> different.  Thus a different E.164 registration model from generic SIP
>> URI provisioning models would neither be a burden on industry nor  create
>> backwards-compatibility issues for the generic solution.
>>
>> Note that if we do not deliver Phase 1 in a very, very timely manner,  
>> the
>> less-than-idea industry solution just might blow up the  possibility for
>> the generic solution, which would be considerably less  than ideal.
>>
>> On Jan 27, 2010, at 1:16 AM, Dean Willis wrote:
>>
>>>
>>> We seem to have an urgent need for a near-term solution for dealing  
>>> with
>>> the subset of the MARTINI problem that, loosely defined, covers  the
>>> "telephone number" problem.
>>>
>>> The typical scenario here is SIP trunking, wherein a SIP Service
>>> Provider (SSP) provides an ISDN-PRI like service over IP to a  
>>> customer's
>>> PBX (or PBXes). In this model, the SSP maps a whole  bunch  of phone
>>> numbers to the PBXes. This mapping is somewhat  dynamic, because the
>>> PBXes usually have DHCP addresses, and don't  have DNS entries that 
>>> point
>>> A records at the PBX in any sort of  particularly useful way (maybe
>>> something like dhcp-10-1-2-3- example.com).
>>>
>>> So what we need is a way for a PBX, when it reboots (or perhaps
>>> periodically) to say to the SSP "Here I Am; You may now send any  calls
>>> for my assigned phone numbers to my current Contact."
>>>
>>> This is indeed a "registration" problem, but it isn't necessarily a
>>> "domain registration problem". The only reason domains get mixed  into
>>> this scenario at all is that we sometimes encode telephone  numbers as
>>> the user parts of SIP URIs, and those SIP URIs have a  domain part
>>> attached to them.
>>>
>>> More interestingly, this problem is a tiny subset of the whole  MARTINI
>>> dynamic binding problem, and it potentially can be solved  much more
>>> simply than the general problem.
>>>
>>> On today's Interim II call, Richard, Adam and I raised the idea of
>>> dividing our problem space, and taking on a near-term milestone for
>>> delivering a MARTINI solution for registering contacts to a proxy,  
>>> where
>>> that proxy maps one or more telephone numbers to the contact.
>>>
>>> We certainly intend to go on and tackle the more general domain-
>>> registration problem, but I (and I believe Adam and Richard) believe
>>> that we get a lot out of tackling the telephone-number subset first.
>>>
>>> In particular, we believe that this approach solves many of the  domain
>>> "responsibility" questions, simplifies the TLS keying/ certification
>>> model immensely, eliminates the cognitive overload of  using REGISTER to
>>> bind a "domain" to a Contact, "route  authentication", and a good many
>>> other open issues that plague our  current drafts.
>>>
>>>
>>> Open questions relating to this approach include:
>>>
>>> 1) Do we deal only with E.164 numbers, or do we need to deal with
>>> private-number "phone contexts"?
>>>
>>> 2) D owe need to explicitly encode the telephone numbers being
>>> registered in the REGISTER request (either directly or through some  
>>> sort
>>> of regexp), or is it adequate to rely on provisioning of the  SSP and 
>>> PBX
>>> (and any UAs) to statically bind the numbers to the PBX  identity, and
>>> then dynamically only register the PBX identity to the  PBX contact?
>>>
>>> 3) Can we "register" with a tel: URI as the "To" value? (this seems  to
>>> require extending RFC 3261)? Note that "tel:" can express both E. 164
>>> numbers and private phone-contexts in one format.
>>>
>>> 4) What happens with GRUUs?
>>>
>>> I believe all of these issues are quite tractable, if we accept the
>>> fundamental idea of splitting the phone-number problem off from the  
>>> more
>>> general domain-registration, and recognize the possibility that  the
>>> solutions to the two problem may reasonably be quite different  from 
>>> each
>>> other.
>>>
>>> So, what do you people think? Is it worth trying to drink the  MARTINI
>>> elephant one sip at a time, or are we going to treat the  whole thing
>>> like a tequila shot and just slam it down?
>>>
>>> -- 
>>> Dean
>>>
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> martini mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/martini
>>
>> _______________________________________________
>> martini mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/martini
> 
> _______________________________________________
> martini mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/martini
> 
_______________________________________________
martini mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/martini