29 jan 2009 kl. 16.36 skrev Scott Lawrence:

>
> On Thu, 2009-01-29 at 15:36 +0100, Johansson Olle E wrote:
>>> Don't use registration.
>>>
>>> Provision the DIDs, then all normal targeting mechanisms work just
>>> fine.
>>>
>>> Someone will say "but what about dynamic addresses" - don't use a
>>> dynamic address for your phone system.
>>>
>> That's a good opinion, but not the reality. There are too many large
>> providers that need a registration and can't place calls on SRV/NAPTR
>> and domains.
>
> Domain names would be a good idea, but they are not required to use
> provisioned connections - you can provision one or more IP addresses
> almost as easily.
>
> Registration was designed to route calls to _a_ contact - it is a
> misapplication of the mechanism to use it for any more than that, and
> that's where the difficulty comes from.
>
> It's a mystery to me why this is so commonly used (yes, I know it is,
> and the sipXecs SIP Trunk support can do it).  Why would I want the
> reliability of my inbound calls to be controlled by my system having  
> to
> 'poll' (that's effectively what a registration is) rather than the
> provider just knowing where to send it?
>

Well this difference between what SIP engineers want to be the
proper way and the way old-time telco people implement SIP is
the huge problem with establishing SIP Connect. It's too far from
reality and is now being pushed even further away.

We need to start where implementations are and lift from there.
Registrations for multiple DIDs does exist out there. I agree that
it's a strange hybrid that really should not be needed, but it's
out. Let's fix that and then move people on.

Right now there's too many ways for service providers to solve
this problem, including home-brewed headers. That doesn't
help anyone.

/O
_______________________________________________
Sip-implementors mailing list
[email protected]
https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors

Reply via email to