Hadriel Kaplan wrote:
>> -----Original Message-----
>> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Paul
>> Kyzivat
>> [snip]
>> IMO that is a value of distinguishing tel from sip/user=phone. With sip,
>> the supplier of the URI determines the routing. With tel, the originator
>> of the call determines the routing.
> 
> You're implying the supplier or originator of the URI cares, or has a way to 
> let its user determine this.  Afaik, there is no button on phones for "let 
> people call me only through sip" vs. "let people call me however they can".  
> Most vendors assume people want the latter, I suspect.  But since most or at 
> least many devices use sip: all the time period, and most proxy/etc. vendors 
> want calls to work because their customers do, they will route it whichever 
> way they can to do so.

As I've said elsewhere it doesn't upset me to have pstn routing used as 
a fallback. But the common behavior we've been discussing has never 
stated the sip routing will be preferentially tried.

Regarding whether the phone gives the user a way to specify this 
behavior: it would be an unusual user who would understand how to use 
such an option. But I can see how particular devices, based on their 
design of configuration, would decide this one way or another. 
Audio-only devices might just use tel all the time, since sip routing 
might not give them any benefit.

Multimedia devices, and devices with fancy audio capabilities might 
choose to advertise a sip URI because there is little benefit to them 
otherwise.

> A second problem is the desired behavior of using the From for later 
> requests.  Since UA's register a sip URI typically, and they use this for the 
> From of their requests, the far-end UAS will get that From.  Since the UAS 
> would then use this URI for its address book to generate new requests later, 
> by your earlier emails this implies that UAS only wants the request to purely 
> stay sip.

This does require some additional clarification. At the moment it is not 
legal to register a tel URI. So you must register the sip equivalent 
instead. There would need to be some clarification about the cases where 
a tel may be used with a sip phone number registration. Or else maybe it 
would be better to make registering a tel URI valid. I don't currently 
have a strong opinion on one vs the other.

> That would be ok if the original request stayed sip the entire path to the 
> UAS,
> and the reverse path sometime later could also stay sip. 
> There are actually cases where that's not going to be true, but let's assume 
> it
> is for now.  So imagine if the original request had a sip From but a tel 
> req-uri.
> For some reason this request gets routed through the PSTN, back over SIP to 
> the
> final dest UAS.  If the PSTN-SIP gateway uses a tel From we're ok, but if it
> uses a sip From then per the desired behavior the UAS would use this sip URL
> in later requests and not be able to use the PSTN or whatever to get it back.

Well, with the semantics we are trying to sort out, it would be entirely 
appropriate for the GW to use a tel uri for To-, From-, and R-URIs. 
Since the incoming call is from the PSTN it has no data to suggest that 
the caller can be reached any other way. I suppose it could use a sip 
URI identifying itself, if it *wants* to take responsibility for any 
return calls, but that is an unlikely scenario.

> That, to me, is why this desired behavior is so brittle.  There are just so
> many times/ways a sip URL in the From or req-uri is used when it ostensibly
> shouldn't, that operators have little choice but to ignore it to make calls 
> work. 
> I think this may be fixable eventually, but I'm not quite sure how to do that
> and keep all the semantics you're looking for, without having calls fail in
> the transition period - which is a non-starter.
> 
> 
>> It is the desire to ignore the domain name in the sip URI that leads to
>> your need to have this special, unenforceable, policy on responsibility.
>> (Well, its enforceable in a legal sense with peering agreements, but
>> that builds a closed garden.)
> 
> People don't "desire" to ignore the domain name, they "desire" to make calls 
> work.  If all UA's sent using tel URLs when they should per your definition, 
> we probably wouldn't be having this conversation. :)

I'm proposing what I think is a more sensible and functional arrangement 
if everybody followed it. Obviously everybody doesn't follow it now, we 
need to figure out if there is a feasible migration path. But its not 
like everything that should work does work today (or we wouldn't be 
having this conversation), so there is need to figure stuff out in any case.

(My general philosophy of architecture is to first figure out when I 
want to be, then figure out how to get there, and finally make 
compromises if I can't find a way to get there. If instead you just find 
the easiest thing you can do, you just end up going in circles and 
making a hodge-podge.)

        Paul
_______________________________________________
Sip mailing list  https://www.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use [EMAIL PROTECTED] for questions on current sip
Use [EMAIL PROTECTED] for new developments on the application of sip

Reply via email to