Internet Telephony and Next Generation Networks

I started being interested in Internet telephony by a circuitous route, as a part of the quality of service research that we conducted at INRIA in the late 80's and early 90's. By 1991, we realized that the new workstations were going to have enough power to perform audio and video compression in software, and that we could conduct actual network experimentation instead of relying only on simulation and modelling. So, with Thierry Turletti, we developed IVS, a software based video-conferencing system that used the H.261 standard for video. (In truth, Thierry did most of the work, I was merely the project leader.) The lessons we learned on IVS helped us contribute to the standardization of RTP and RTCP in the IETF.

The Internet Telephony conducted at Telcordia led to the development of the "Call Agent Architecture" that enables very large scale configuration, moving Internet telephony into the main stream of telecommunications. One of the most visible outcome of this work has been the development of new standards for Internet Telephony:

The Simple Gateway Control Protocol was defined in the spring of 1998, and published as an Internet draft. There is more information about SGCP on SGCP pages published by Telcordia's Applied Research lab.
SGCP was merged with the IPDC proposal sponsored by Level3 Communications. This led to the definition of the Media Gateway Control Protocol, MGCP, which has been adopted as a standard by the Cable Labs. We submitted the first draft (V0.1) of MGCP to the IETF in October 1998, followed with updates in November 1998, January 1999, and February 1999. The firt "official" version of MGCP is defined in RFC 2705.
Further standardization effort has been pursued in the IETF, in the MEGACO working group, and also in the ITU/SG16, under the code name H.GCP. I am one of the author of RFC 3015 that defines the MEGACO protocol (also H.248), but I have mixed feelings about it: there are way too many compromises in this design; we would have been better off simply upgrading MGCP.

Since joining Microsoft, I have shifted my interests away from the server based "call agent" architecture, back towards a more "peer-to-peer" architecture, that is using SIP rather than MGCP or Megaco: the call agent architecture is a fine way to control dumb devices, such as telephone line terminations in an Internet telephony gateway, but it is not the right solution if the device is "intellignet", e.g. a PC or a PDA. With several colleagues, I have proposed extending the session invitation protocol (SIP) to cover "presence" and "instant messaging"; this is currently a ork in progress, in the SIMPLE working group of the IETF. A preliminary version of the SIMPLE extensions to SIP is implemented in Windows Messenger.