[asterisk-dev] Kill the user, kill the user!

[Johansson Olle E]

10 dec 2007 kl. 10.23 skrev Klaus Darilion:

> Johansson Olle E wrote:
>> 7 dec 2007 kl. 17.45 skrev Igor A. Goncharovsky:
>>
>>> Hi!
>>>
>>> Johansson Olle E wrote:
>>>> Incoming calls are handled this way
>>>>
>>>> * First, we match on peer object name with the From username
>>>> * Then we try to match on IP/Port
>>>> * If we can't match, we send to the context defined in the  
>>>> "general"
>>>> section in sip.conf or to "default".
>>>>
>>>> If you can find any way this may give you problems, please inform  
>>>> me
>>>> now. Otherwise, I'm going
>>>> to test this branch with all of you. I don't foresee any problems
>>>> going ahead with this in trunk.
>>>>
>>>> Feedback is appreciated!
>>>>
>>> Should it solve my problems with incoming calls from my telco. I  
>>> have
>>> registered many lines on single SBC then incoming call have the same
>>> IP
>>> and port. Difference only in To: user name. Incoming INVITE have not
>>> auth, but when we send final BYE we have problem with auth name,
>>> because
>>> it is wrong selected on initial INVITE. Can also peer search able by
>>> To
>>> field?
>>
>> Well, that is a different thing. After I've removed the user, my plan
>> was to add "service" and "trunk".
>
> What if Asterisk uses a dedicated Contact URI when registering to the
> service provider. Then, incoming requests can be distinguished based  
> on
> the RURI. (this is what eyebeam and SNOM does by adding an instance/ 
> line
> parameter to the contact URI).
>
> Thus:
> 1. check RURI if it matches an existing registration (optionally  
> verfiy
> if requests comes from the specified IP address as additionally  
> security
> check)
>
> 2. check From URI. If the domain correlates with the local configured
> domains then challenge the user (realm=domain) (old user behavior)
>
> 3. check source IP (optional port) (old peer behavior). Optional do
> authentication
>
> 4. use default context

A very good idea! And something I've been working on for a while.

Already have that code. In fact, it was part of test-this-branch.

Working on integrating it into trunk.

Thanks for the feedback!

/O