I don't mean to hijack this conversation, only add a data point...
I have a great deal of respect for the people who have done the heavy
lifting in BEHAVE, but it seems like every time we meet, someone discovers a
new and previously un-observed NAT behavior that Is Not Helpful. This week
was the best yet...
In a recent posting
(http://list.sipfoundry.org/archive/ietf-behave/msg01189.html), Dan Wing
said:
That isn't quite the scenario. The scenario where the UDP/TCP
interworking is useful is this:
Alice---[NAT/firewall]---[TURN]------------[NAT]---Bob
<-------TCP---------------->|<-----------UDP-------->
Where Alice's NAT/firewall device blocks UDP. Many company
firewalls have that behavior, which cause applications such as
Skype and Yahoo Voice to send their traffic over TCP.
Bob, on the other hand, has a 'normal' NAT, however Bob's
endpoint has no support for framing RTP over TCP. This is
pretty common -- many endpoints have no ability to send their
RTP traffic over TCP.
So, this functionality provides a way for Alice and Bob to
communicate where they couldn't communicate if they were both
forced to use UDP (because Alice's firewall blocks UDP).
OTOH we definetly need to nail down framing in the TCP->TURN->UDP
scenario. I think we can use the TCP framing mechanism that Jonthan
proposed at the behave meeting; the server would emit every
TCP 'chunk' as a UDP packet.
I'm still thinking about the "telling our children about the ancient past
when it was possible/legal for an ordinary person to put a server on the
Internet" remark from Thursday's plenary. Dan's note, quoted above, is an
excellent summary of where we are now, and it is not too much of a leap to
JohnK's favorite rhetorical question, "where are we going, and how did we
get in this handbasket?"
So my point was, I'd really like to take a chance on some IAB statements
about things that need to be stated about our architecture. They might be
ignored. Would the result be any worse?
Thanks,
Spence
P.s. And how many working group meetings did YOU sit in this past week,
where firewall and NAT traversal were affecting our protocol designs? I
think I counted six different working groups (including BEHAVE). Dan Wing
could be thinking about USEFUL problems, if we weren't distracting him with
stuff like this. Ditto the rest of the ICE/TURN/STUN crew.
_______________________________________________
Ietf@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ietf