Re: [AVT] Re: RTP/RTCP Port Sharing -> SCTP

Michael Thomas wrote:
> 
> Jonathan Rosenberg writes:
>  > Then you read draft-rosenberg-sip-entfw-02.txt, due out this week, which
>  > explains how to handle this case.... :) The mechanism in there uses
>  > bidirectional streams (which we call symmetric) whenever only one side its
>  > natted, and otherwise uses a network intermediary outside of the firewall
>  > when both are behind nats that don't support Christian's
>  > http://search.ietf.org/internet-drafts/draft-huitema-natreq4udp-00.txt. When
>  > contacting this network intermediary, its still useful to use symmetric RTP.
> 
>    Sigh. If the end result of most media sessions is a rendezvous
>    to a intermediary, then our bragging rights over the IN will
>    be nullified. I know that this is the implication of NAT's, but
>    it still depresses me.

It would be nice if all the effort expended on fixing the NAT stuff
would be matched by people deploying IPv6 applications (enabling IPv6 on
recent Solaris and Linux boxes, as well as routers, seems pretty
trivial, from our experience) or dispelling the notion that you can't
get addresses, which seems to often be perpetuated by people who have
other agendas (such as making it difficult to use more than one computer
on a home DSL/cable modem).

That said, I wonder if it wouldn't be easier to just have the client set
up one secure tunnel to this outside entity with a real IP address, for
all applications that are stymied by NATs, rather than inventing a new
mechanism for each application. This does mean two relaying points, but
avoids all the negotiation and myriad special cases. Can this be done
with the typical VPN setups in some of the more recent OSs?

-- 
Henning Schulzrinne   http://www.cs.columbia.edu/~hgs

_______________________________________________
Audio/Video Transport Working Group
avt@ietf.org
http://www.ietf.org/mailman/listinfo/avt