Re: another ListeningPoint issue ...

Ryan,

With regard to the last sentence in your post, I seem to recall some 
discussion about having to switch to use TCP if a message that's being 
sent is too big for UDP. I thought that might be one of the reasons to 
have multiple listening points under a SipProvider. (but I suppose both 
sides would have to do it to make it work in the case you ask about). 

To anyone on the list:

Is there a particular way that listening points should be grouped under 
SipProviders of a given stack? (other than the rule of at most one 
listening point of a given transport type per SipProvider). IE: Given a 
user-specified configuration containing any number of 
IPaddress/port/transport combinations for a single stack/SipListener 
instance, I have created a separate SipProvider for each IPaddress/port 
combination and grouped the LPs for the different protocols underneath 
those. From an application perspective, is there any reason why you 
might NOT want to do that?

Thanks for any advice-
Becky

Ryan Mitchell wrote:

>Using the following setup:
>
>SipProvider A  --  ListeningPoint A : 192.168.2.3, 5060, udp
>SipProvider B  --  ListeningPoint B : 10.0.2.3, 5060, tcp
>SipProvider C  --  ListeningPoint C : 10.0.2.3, 5060, udp
>
>
>All under one SipStack, all using the same SipListener implementation.
>
>I receive a REGISTER request from C.  Next create a ServerTransaction
>st = sipProvider.getNewServerTransaction(req);
>
>... application processes the request ...
>
>Response resp = msgFact.createResponse(Response.OK, req);
>and send it: st.sendResponse(resp);
>
>Problem is response gets sent out A.  The reason I think is because the
>SIPServerTransaction code gets the message channel to use from
>sipStack.createRawMessageChannel(this.getPort(), hop); which finds the
>first MessageProcessor matching the port number and hop protocol.  And
>in my case it finds A (udp, 5060), whereas C should be the correct choice.
>
>If I reconfigure A to use port 5062, no problem.  Is this a problem in
>the stack or am I missing something?  I think the only requirement per
>RFC 3261 is that is the request is received "secure" that the reply is
>also sent secure.  That said, are there any cases where you would want
>to respond to a request on a different listening point from which it was
>received?
>
>thanks,
>--Ryan
>
>
>  
>