Re: [MMUSIC] Requiring support for nonconformant intermediate systems (Re: FW: New Version Notification for draft-ejzak-mmusic-bundle-alternatives-00.txt)

[[e-mail] (Dale R. Worley)]

> From: Harald Alvestrand <[e-mail]>
> 
> Designing around intermediaries that pass along information that they do 
> not understand, and then expect to act independently based on the 
> message that they partially understood, seems to me to be a quite 
> limiting approach.

If we're going to succeed, we've got to be aware of the environment in
which the solution operates.  In principle, intermediate systems
should operate transparently, but they don't.

> We discussed this at the interim (I think Cullen was pretty vocal that 
> there's a limit to how much brokenness we can design around, and he 
> thought intermediaries passing BUNDLE without understanding BUNDLE were 
> clearly beyond that limit).

Maybe Cullen was vocal about that, but he also points out that C***o
NATs *have* to pay attention to the port numbers in media
descriptions, so they enforce the uniqueness of port assignments.

Because of problems like this, four of the "compatibility desiderata"
I included in my proposal document the behavior of middleboxes we have
to tolerate.
(http://tools.ietf.org/html/draft-worley-sdp-bundle-02#section-3.2)

Dale