On Wed, 2009-01-07 at 11:02 -0500, Paul Kyzivat wrote:
> What Scott says seems reasonable. In his case, if the feature is turned
> off it really is a proxy. If the option is turned on, and SDP is
> updated, then the best you can say is that it is a proxy with
> non-compliant behavior.
... and I'm tarnishing my reputation as a Purist by releasing it :-)
But the Real World is like that... the feature is needed, even if it
doesn't fit into the nice categories.
Incidentally, using that NAT traversal feature unavoidably creates
problems for some things like redundant proxies that would theoretically
be possible without NATs, but our experience so far has been that:
* Most deployments have to deal with NATs somewhere. We chose to
do as much compensation for this as we could by modifying SDP in
the "proxy" rather than a full B2BUA (the distinction being that
there is not a new set of dialog identifiers - both sides are
using the same call-id and tags).
* In interop testing (some time ago - this may be improving),
we've found that many phones don't do the right thing with an
SRV name in a Route, so at present the Record-Route inserted by
the proxy is always an IP address, which makes the theoretical
redundancy for onging calls moot anyway.
There are similar annoying issues with registrations, etc that have had
lots of discussion...
_______________________________________________
Sip-implementors mailing list
Sip-implementors@lists.cs.columbia.edu
https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors
