Diego Serafin wrote: > 2006/7/24, Pekka Pessi <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>>: > > On 7/20/06, adi baron <[EMAIL PROTECTED] > <mailto:[EMAIL PROTECTED]>> wrote: > > Can any one tell me if it's convenient to modify the application > so that > > sofia-sip could support multiple Endpoints? > > > > How the multiple endpoints should be supported? I fyou are not > > planning to do anything fancy, you can just register multiple AoRs > > (address-of-record). If you don't need to register , you can also > > examine the request URI in incoming initial requests. > > > > - What do you mean by "doing anything fancy" ? > > The problem with multiple endpoints is to determine correct endpoint > for an incoming SIP request. Usually that is done with the > request-URI, e.g., the userpart of the request-URI contains the name > of the endpoint. > > > - What can not be implementing using this method (of sharing the > same nua_t > > object) ? > > If you want to register the multiple endpoints, your application must > specify each request with an AoR and an unique Contact header. The > "fancy" things refer to situations where the SIP stack (nua) is > expected to generate the Contact header. For instance, if there is a > NAT box between your application and the proxy or the rest of the > Internet. > > Currently, nua generates a Contact header with an empty user part. > When your application registers two endpoints with different AoRs > using the same Contact, it might be impossible to say to which > endpoint it should direct an incoming request. That is because the > only information about the intended target is the request URI, which > is taken from the endpoint's registered Contact header. > > > - while reviewing the code i have noticed states for call hold > and active > > call are maintained. > > cant it be a problem while the object now responsible for multiple > > endpoints ? > > Nope, they should be call-specific. > > > What would it be your suggestion about implementing a multi-endpoint > architecture with sofia-sip? We are considering it as a replacement of > our current stack for an IP-PBX product and we currently use a > multiple-UA approach (i.e. one SIP 'channel' to one SIP stack). as we > do for ISDN (and analog) lines control but are considering a single UA > with a sort of multiplexer acting on a multitude of contexts. > ------------------------------------------------------------------------ We looked at the multiple stack approach, but turned it down since each UA would have to use a different UDP/TCP port. Also, the memory foot print got to be too big for our application. The single UA approach requires multiplexing on the URI and adding multiple AORs, and appears to function well for us with 10s of endpoints. I have not tried 100s of endpoints, but assuming you have the CPU to support it, I would expect it to be ok too.
------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Sofia-sip-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/sofia-sip-devel
