On 2007-6-1, at 21:09, ext Dan Wing wrote:
We need a document that consolidates NAT keepalive mechanisms for
UDP-based protocols, most notably RTP (and, to a lesser extent,
SIP). Currently this is spread in various documents and there
is no consolidated guidance to implementtors on the strength
or weakness of various approaches.
I'm wondering if it'd make sense to add this guidance as a
section in
http://tools.ietf.org/html/draft-ietf-tsvwg-udp-guidelines?
I agree "send periodic UDP traffic" would be good in that
document. REQ-6 of draft-ford-behave-app-05 may provide the
beginnings of some useful text (CC'ing the authors of that
document).
As with all applications on today's Internet, an RTP
application needs to be robust and able to deal with whatever
is sent to it, so perhaps you're right -- an application-
specific keepalive document may well be overly specific and
we should, rather, document what any UDP-based application
should do to prevent middleboxes (such as NATs) from
clobbering a flow.