Dealing with Varied Device Capabilities in SIP

[Ben Teitelbaum <>]

Ken Klingenstein and I were chatting yesterday about the following
problem.

  How does a caller reach a recipient on a device having capabilities
  preferred by the caller?

For example, suppose Bob has a single identity 
(sip:)
 and
has a variety of devices registered under it, including a voice-only
phone and video-enabled terminal.  Further assume that Bob's proxy
will fork to all of Bob's registered devices.  Now suppose, Alice
wants to have a video call with Bob.  

There are several questions here:

  1) How does Alice ring Bob's video phone?
  2) How does Alice ring *only* Bob's video phone?
  3) How does Alice know a priori whether she could reach Bob by
     video?

SIP experts out there, please help me understand how this is supposed
to work and where, if anywhere, there are missing pieces.

Taking these one at a time...

  1) How does Alice ring Bob's video phone?

Can the SDP in Alice's INVITE require certain video capabilities or
are SDP's media attributes just a big smorgasbord, for which a
recipient device need only match one media type to ring?  For
example, assume that the SDP in Alice's INVITE advertises PCMU/8000
(G.711), then will all of Bob's devices will ring (since they are all
capable of G.711)?  If so, Bob will probably pick up his phone and
Alice will have to tell him that she really wanted a video call and
call back.  My hunch is that with standard SIP and simple forking,
this is what would happen.

  2) How does Alice ring *only* Bob's video phone

SIP seems to have an answer to this in the form of the SIP Caller
Preference Extension.  This allows clients to advertise the
capabilities of terminals on registration and allows callers to
express their preferences in an INVITE message (through new headers,
not SDP).  A SIP proxy can then merge the set of caller preferences
and capabilities with those of the callee to route the call
appropriately.

See the first seven or so slides of 

http://www.jdrosen.net/papers/sipintro_summit2001.ppt

or

http://www.ietf.org/internet-drafts/draft-ietf-sip-callerprefs-08.txt
for the whole story.

Finally...

  3) How does Alice know a priori whether she could reach Bob by
     video?

Do SIP or SIMPLE have an answer to this?  Is there a standard means
for a presentity to publish its device capabilities?  Can anyone point
me to the right RFCs or IDs?  Is anyone aware of a demonstrable
implementation of this?

In the an H.323 context, the Internet2 middleware folks have proposed
an LDAP schema called "commObject" to publish device capabilities in a
directory.  See:

http://middleware.internet2.edu/video/draftdocs/draft-nmi-edit-vidmid_vc-commObject_White_Paper-1.0.html

Offhand, it would seem to me that the set of devices at which a user
is reachable is a highly-dynamic set.  Therefore, a presence approach
to this problem would seem preferable to a directory approach.  How
might the commObject work be useful in a SIP/SIMPLE environment?  Any
thoughts?

-- ben

---------------------------------------------------------------wg-voip-+
For list utilities, archives, subscribe, unsubscribe, etc. please visit the
ListProc web interface at 

    http://archives.internet2.edu/

---------------------------------------------------------------wg-voip--