Tom,
Thanks for the nice summary of events regarding gateway decomposition. I just want to comment on a few minor things (see below.)
--Hong
Tom-PT Taylor wrote:
Here's a brief history, from my point of view -- others are free to add their viewpoints.
April 1998: Bellcore announces SGCP (Simple Gateway Control Protocol). They have been prototyping for several months. Hong Liu was one of the participants. I have the impression that many of the ideas came from Christian Huitema, including the use of SDP, the separation into local and remote descriptors which are effectively negotiated between MGs, and the use of application-level framing instead of a reliable transport layer. SGCP meets with a hostile reception in SG 16 and TIPHON because of its centralizing architecture, which differs subtly from that of H.323.
May 1998: A cluster of activity around the issue of control of VoIP gateways where SS7 is terminated on separate signalling gateways (to conserve point codes). Nortel is proposing the use of an extension of DIAMETER, a protocol under development at the IETF, to do device control. Ascend and Bay have extensions to Q.931 for the purpose, which they eventually rationalize into a single protocol proposal.
May 1998: Level 3 announces the formation of a multi-company Technical Advisory Committee to work on an IP Device Control protocol (IPDC). The TAC is presented with a 100-page specification previously developed by Ascend and XCOM as their starting point. Other companies have in-house protocols, and SGCP is also considered, but the TAC ends up using a modified IPDC over a DIAMETER base. IPDC includes the important concepts of event packages and scripts.
June 1998: The TAC presents their list of requirements to the H.323 Rapporteurs' meeting in Cannes. Intel presents an abstract proposal for H.323 gateway decomposition. Bellcore presents SGCP. Work on gateway decomposition does get started, despite initial hostility and suspicion. A TIPHON meeting around this time at least became familiar with the topic.
I recalled that SGCP was discussed in Cannes. But SGCP as a protocol was not submitted to ITU-T SG 16 as a contribution. In fact, only MGCP was submitted to ITU-T SG 16, and the first submission was in 11/98 at the Torino's meeting.
August 1998: the Level 3 TAC concludes its work and presents a first draft of IPDC to the IETF and SG 16. All the discussion of SS7 gateways and remote control of VoIP gateways culminates in a Birds of Feather (BOF) session at the IETF. Out of that eventually comes the decision to form two working groups: SIGTRAN (transport of SCN signalling over IP) and MEGACO (Media Gateway Control). That decision doesn't come until December, but in the meantime there is active discussion of issues on the IETF lists. TIPHON adds gateway decomposition to its architecture, but is having trouble getting a firm view of that architecture.
September 1998: SG 16 considers gateway decomposition and commits itself to working on the topic. Terms of reference are drawn up, along with a reworking of architectural views initially drawn up in June. Bellcore and Level 3 announce a merger of SGCP and IPDC into MGCP. The latter is mostly SGCP with the addition of event packages and termination wildcarding syntax.
The actual merger of SGCP and IPDC occurred around 10/98. MGCP was announced at the end of 10/98.
November 1998: TIPHON architectural views are settling down. Lucent brings in MDCP (Media Device Control Protocol) as a hand-written draft to the SG 16 Rapporteurs' meeting in Torino. This is completely different from MGCP, particularly in its connection model.
December 1998: first meeting of the Megaco working group at the IETF. A first draft of a requirements document comes out shortly thereafter.
March 1998: a bitter debate over connection model at the Monterey Rapporteurs' meeting. In a procedure unusual for SG 16, the MDCP connection model is chosen via an indicative poll.
That was Feb 1999. I would not forget this, -:)
April 1998: compromise reached at last on the connection model, at the IETF. The new model accommodates all of the additional requirements which have arisen beyond the original design premises of SGCP, and improves on both the SGCP/MGCP and MDCP models.
That was March 1999 IETF meeting in Minneapolis when the compromised model was reached by a design team. The actual I-D came out in April 1999.
May 1998: formal Megaco/H.GCP cooperative arrangement announced.
Again, it was May 1999 ITU-T SG 16 meeting in Santiago, Chile. And the IETF/ITU-T arrangement was announced (prematurely) in a very strange way, and the announcement was to the surprise of most people in the meeting.
-----Original Message----- From: Derks, Frank [SMTP:F.Derks@PBC.BE.PHILIPS.COM] Sent: Wednesday, June 30, 1999 6:24 AM To: ITU-SG16@MAILBAG.INTEL.COM Subject: Re: MGCP reference (was Re: Megaco/H.gcp conference call)
HI Nancy,
maybe it is a good idea to post the history of all these protocols and their relations (if any) to this list?
Frank
-----Original Message----- From: Nancy-M Greene [mailto:ngreene@NORTELNETWORKS.COM] Sent: 29 June 1999 21:34 To: ITU-SG16@MAILBAG.INTEL.COM Subject: Re: MGCP reference (was Re: Megaco/H.gcp conference call)
Chip, you are right, the name Telcordia/Level 3/RSL COM/Vertical MGCP is not convenient. Christian has already made the same comment you have. I said in reply to him, and to the list, that I did make an error in adding Cisco to the qualifier for MGCP. Since then, I refer to the new Megaco protocol as the Megaco protocol, or the Megaco/H.gcp protocol, and to MGCP as MGCP.
Since a new version of MGCP has come out, version 1, (draft-huitema-megaco-mgcp-v1-00.txt), this distinction is even more necessary. It makes sense that megaco is part of the file name, because there are parts of MGCP that still need to be turned into RFCs for packages for the Megaco/H.gcp protocol, but it means the new Megaco/H.gcp protocol can never really be called MGCP - it is too confusing. It would be good at some point to pull the packages out of MGCP to stand alone as a separate document for Megaco. And perhaps the UDP + timers part of MGCP so that it can stand beside TUDP, MDTP, and TCP as alternatives for Megaco transport.
Nancy
Nancy M. Greene Internet & Service Provider Networks, Nortel Networks T:514-271-7221 (internal:ESN853-1077) E:ngreene@nortelnetworks.com
From: Chip Sharp[SMTP:chsharp@cisco.com] Sent: Tuesday, June 29, 1999 2:53 PM To: Greene, Nancy-M [CAR:5N10:EXCH]; ITU-SG16@mailbag.cps.intel.com; megaco@standards.nortelnetworks.com Subject: MGCP reference (was Re: Megaco/H.gcp conference call)
Once or twice is a mistake. When the mistake is continually repeated I wonder...
Cisco is NOT an author of MGCP. Referring to MGCP as "Telcordia/Cisco MGCP" denigrates the hard work of the many people that actually did author the draft as well as the many people and companies that contributed. Maybe it isn't convenient to call it Telcordia/Level 3/RSL COM/Vertical/ MGCP?
Chip
At 06:44 PM 6/16/99 -0400, Nancy-M Greene wrote: ...snip...
mechanism. We need to choose one. There are two proposals on the table:
- the UDP + timers method in the Telcordia/Cisco MGCP -
draft-huitema-megaco-mgcp-v0r1-05.txt, and
...snip...
Chip Sharp voice: +1 (919) 851-2085 Cisco Systems Consulting Eng. - Telco Reality - Love it or Leave it.