[OpenSER-Devel] SF.net SVN: openser: [4362] trunk/modules

[Dan Pascu]

On Wednesday 11 June 2008, Bogdan-Andrei Iancu wrote:
> regarding client_nat_test - the nat test is also based on SDP IP, which
> is can be provided via MI, so , it can be private (integration with
> some AS).

I did not include the SDP checking functionality in this function (this is 
the only thing left off from the extra stuff supported by the nathelper 
version) because it has design issues which I would rather see discussed 
after the release to find a better alternative.

The issue with checking the SDP IP is that it is an inconsistent and 
unreliable test for many reasons. The nathelper implementation checks the 
first c= line in the SDP body. There are multiple issues with this:

1. SDP may contain multiple sessions. What if 2 sessions, one with a 
private IP the other with a public IP?
2. First c= line may be a session level IP of a stream level IP
3. The c= line may point to a different IP than the one from the device 
that generated the signaling
4. The first c= line may be from an unsupported stream that still has the 
original address, or from a supported stream that already had its address 
replaced by mediaproxy. This depends on the stream order an is not known 
in advance.

-- 
Dan