[SR-Users] [sr-dev] freezing for major release 3.1

[Timo Reimann]

Iñaki Baz Castillo wrote:
> 2010/7/21 Carsten Bock <lists at bock.info>:
>> Hi,
>> it is rather simple:
>>
>> The command specification (http://rtpproxy.org/wiki/RTPproxy/Protocol)
>> states the following:
>>
>> Update session, creating a new one if it doesn't exist.
>> U[args] callid addr port from_tag [to_tag [notify_socket [notify_args]]]
>>
>> The NATHelper Module currently does not send the optional notify
>> socket to the RTPProxy. The enhancement makes this "notify_socket"
>> configurable and will send it to the RTP-Proxy, if set. On the other
>> end, i have modified the RTP-Proxy to check, if it is an HTTP-Socket
>> (XML-RPC) or an Unix-Socket. If it is a Unix-Socket, it works as it
>> does now. If the Socket starts with "http://", it will query the Proxy
>> by executing "dlg_list" with the "Call-ID" and the "From-Tag". If the
>> dialog is found, it will call "dlg_end_dlg" with the according
>> parameters...
> 
> It's great :)
> 
> However take into account that "dlg_end_dlg" just works for confirmed
> dialogs (state 4).
> There are new fixes about it by Timo so it doesn't leak when running
> "dlg_end_dlg" for an in-progress dialog.
> Also I think that a new commit from Time makes "dlg_end_dlg" just to
> work in case the dialog state is 4 so nothing would be required from
> the Rtpproxy.
> Terminating in-progress dialog is still not implemented (but planned AFAIK).

All correct; little addendum: Proxy-initiated dialog termination works
both in state 3 (2xx response received but ACK request from UAC yet
missing) and 4 (2xx response and ACK request received) because the
dialog module stores caller data already in state 3. I neglected that
detail in the commit message as both states are denoted "committed" in
Kamailio but it doesn't make much of a difference.


Cheers,

--Timo