[MMUSIC] Quick review of draft-rosenberg-mmusic-ice-nonsip-00

["Frank W. Miller" <fwmiller at cornfed.com>]

Greetings,

I decided to do a quick review of the document.  Here are my
comments/questions.

1) Overall, the document is structured to reference ICE heavily.  I wonder
whether it would be good to bring in selected portions of ICE to this
document to make going about an implementation easier?  My reasoning is that
there may potentially be a lot of applications that are not SIP related and
by reducing references to ICE, you may reduce the pain of someone having to
deal with understanding SIP to get this to work.

For example, the way it's written now, you basically have to go read ICE
before reading this document.  Wouldn't it be great if I could read this
document and implement to it without having to refer to ICE?

2) An protocol ladder diagram for at least the nominal case would be good.
I know its simple but what I'm really looking for is something that includes
the overall picture, i.e. use of STUN, use of NICE, possibly use of TURN all
wrapped up in a single ladder.

3) Some example messages would be great!  An example of Initiate and Accept
messages would be helpful.  My reasoning here is that since you're
substituting MIME for SDP syntax, some more specifics on the use of the
spec'd BNF would be helpful for implementers.

4) Along those lines, while the BNF for the MIME portion is spec'd the rest
of these messages is not.  Do the Initate and Accept messages have some
header portion like a SIP message with a MIME body?  It wasn't clear from
the text how they are supposed to be structured.

5) It would appear that the subsetting you are doing here really boils this
down to hole-punching for a single connection with the ability to use a
relay server if necessary.  Should there be some reference to the previous
hole-punching work?

6) So now the obvious question, which I'm sure you're anticipating and which
is likely more philosophical.  Why could this protocol not be used with SIP
implementations as well?  I know we have ICE itself but...  I suppose if you
used this for each "connection" in a SIP session the result would be even
chattier than ICE is now but this is certainly somewhat easier to implement
given all the subsetting.  While I have a feeling I know what you think
about this, I'd still be interested in hearing your rationale.

Thanks,
FM


_______________________________________________
mmusic mailing list
mmusic at ietf.org
https://www.ietf.org/mailman/listinfo/mmusic