> Damian wrote:

> > Why don't we fix the problem where it really is: proxy and 
> > registrar should read dial plan when it changes and start 
> > using new dial plan when it makes sense for them: they know 
> > after all best when it's safe to do that.

Paul Mossman wrote:

> Not sure how difficult it would be to coordinate that in current
> architecture.  But it would probably be considerably easier if the proxy
> and registrar were unified into a single component......

Moderate difficulty.  

We'd have to rewrite some of the code that processes the various
configuration files, but some of that code desperately needs rewriting
anyway, and there are some features pending that would overlap nicely
with that work.  There would be significant opportunity for efficiency
improvements while we were at it - the current processing of the
mapping/fallback/auth rules is "suboptimal".

As for signaling the change, most of the links in the chain are already
there - because replication happens through sipXsupervisor, and it knows
which processes need which files, it knows which processes to signal.
We would certainly need to add a standard mechanism to pass that
indication to the process (there are a few alternatives that have been
discussed already).  I suspect that we'd want to add an indication from
sipXconfig that reconfiguration is complete so that when it needs to
update multiple files (as it does for dialing rules) the process would
not be signaled to reconfigure until all of them were done - this could
either be a new signal, or we could just add this semantic to the
updating of the configuration version stamp (that is, no process is told
to reconfigure by sipXsupervisor until the configuration version stamp
is written, even if its value has not changed).


_______________________________________________
sipx-dev mailing list
[email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-dev
Unsubscribe: http://list.sipfoundry.org/mailman/listinfo/sipx-dev

Reply via email to