[MMUSIC] SDP attribute level for disaggregated media [was Re: 4566bis outstanding issues]

[Marc Petit-Huguenin <[e-mail]>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 06/06/2014 03:19 PM, Ali C. Begen (abegen) wrote:
> Hi everyone
> 

[...]

> 
> 
> These are the issues I was able to identify. There is one more thing
> related to ICE and I sent an email about this to Marc Petit-Huguenin. As
> soon as he responds and depending on his answer, I might add one more
> issue.
> 

Let me first explain the problem as I perceive it:

The SDP was designed so that different endpoints can be part of an RTP
session, i.e. each m= section can used a different IP address.  For example
one can imagine that during an IETF meeting each microphone has its own IP
address and stream audio directly to some destination (in that example the IP
address in the SDP is only use for RTCP).  If this use case was not possible,
then it would be only possible to use c= at the session level.
draft-loreto-splices-disaggregated-media was, I think, the first document to
call this "media disaggregation".

Now if we expand this to ICE, it means that each ICE agent declared in a SDP
has to be independent of each other, i.e. each microphone in our example has
it's own ICE agent, and they do not even to be from the same vendor.  That
means that all SDP attributes needs to be at least media-level attributes, and
eventually both media- and session- level attributes.  IANA fixed the problem
for ice-mismatch, but the problem still exists for ice-options, and fixing it
is part of rfc5245bis.

Now my suggestion was that for future SDP attributes, people should really
think about the disaggregated case when they choose between a session-level
only attribute or an attribute that works at both session- and media-level,
because using a session-level only attribute prevents media disaggregation.  I
could not find this problem in any existing attribute at the time I did the
analysis but in ICE attributes, but note that this is not a problem specific
to ICE.

So I would suggest adding some text after this paragraph in section 8.2.4:

'Submitters of registrations should ensure that the specification is
 in the spirit of SDP attributes, most notably that the attribute is
 platform independent in the sense that it makes no implicit
 assumptions about operating systems and does not name specific pieces
 of software in a manner that might inhibit interoperability.'

Something like this:

'Submitters of registrations should also carefully choose the type of
attribute. They should not choose a session-level only type when the attribute
can have different values when media is disaggregated, i.e. when each m=
section has its own IP address on a different endpoint.  In that case the
attribute type chosen should be "both".'

Thanks.

- -- 
Marc Petit-Huguenin
Email: [e-mail]
Blog: http://blog.marc.petit-huguenin.org
Profile: http://www.linkedin.com/in/petithug
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQIcBAEBCAAGBQJTngvsAAoJECnERZXWan7EkoEP+QH/FamXyYcisXxR8vtRBUAV
D0MrwvRzzjQfAEW2aRSYPC12de0UeC7uYCOx9l7Cv1NKcFccdD/3NsDUNYy5C++Q
siL0YAcTIhuCiBYkYPUEq3Ix3CrXGnspBV2kbJdcBPgGuIw7jXNHQh5l5bvoAtJs
isHGev6rn6wGR2WheiYtlQ1AInFmjD5MqOyu9xtAAbdIJBWrqFzDsSq4vAevSyzL
EYPgW6Op6UoP6zGIB45PRXiGP//5Gj8jri9wXjp3M8cyvgpK6CoHoNe6m1rGuWqS
pMOh6vZRjRbkRpcufJ29tuV8SJ513m5x0pDSiCZAyNi47q3zgN+W++i4eX1ujnbN
D/UjmQmacSFO3N/QWvuAlZH9VRSVHdUbyw/NXUFFhCjexSokn/nBh7eFFvwMJyp6
ZuxpKjcg+wrOxjkocSq8bNXcQLR1yohADKN1ePyHHvZ84+gkinlIpKMuegEBBFi5
OVe0O4QiRjKd3h1ewO98tzPUKVhaD9nSkLT7eOjHU3Xfz5Ol5tlA50+5P7tcN43h
OO10nUijN4lxBsTi0WNC2sFpE3WXQ8gAkWHKyedfeMx9Uvl2SF18QFujAeD1oYIN
jS+SuZOlO7S6t2naMzyJj10yOxzgSnNYBGBQQ2yP1oE1qtiZjCDhTTaJIMjTgCYu
V/fj5tQ5v43FLPpKlD+8
=qu35
-----END PGP SIGNATURE-----