Re: [reSIProcate] Setting via headers for STUN

[Alan Hawrylyshen <alan@xxxxxxxxxxxx>]

On 2006.06.03, at 14:18 , Matthias Moetje - TERASENS GmbH wrote:

Cullen,

 

thanks for your comments. Admittedly in some cases it

might just not be required to change the via headers,

but you can't expect this behaviour from all other parties.

 

Matthias - SIP provides for this with the received= parameter on Vias. Almost all modern SIP implementations support this.

Setting the Via to the STUN discovered IP is not correct.

And I can't see why it should be wrong to do so.

Some popular SIP phones we have examined are doing

exactly the same thing.

"Everyone else does it" is not a good reason for us to do something, especially when there are potential problems in doing so.

:-)

Furthermore I remember that we have been doing a

test with a VoIP provider where the connection didn't

work without changing the via header.

 

Anyway, the problem has been solved through the new

ability to add a message decorator in dum which can

change the via headers after transport selection, so

anyone can decide wether to change the via headers

in a STUN scenario or not.

I'd say, doing so will increase the chances of successfully

establishing a connection with an arbitrary SIP

implementation in a STUN scenario.

I think you will find some resistance to making this the default behavior. Please consider doing this locally only or we should have a fairly open discussion around this. I fear other list members will object. I object mildly - we have had a general philosophy of not doing things in reSIProcate that are against the standard. At the same time, the stack should provide enough flexibility for a local user to do what every they like with the stack.

Thanks

Alan