I had the opportunity for the past several years to work in a very large working group in the W3C and also participate on some different IETF activities. I’ve largely stopped working on formal standards work in the past 9 months - although there is one RFC draft that we’re close to completing related to SIP and VoiceXML.

Yesterday, after over 4 years of work, the W3C recommended VoiceXML 2.1. This might seem reasonable to anyone involved in standards activities. Let me put this in perspective. The 2.1 spec was a minor, orthangonal “maintenance” update from the VoiceXML 2.0 specification (which also took many years to finalize). The goal was something we could get through fast and something that (a) a majority of participating vendors had already implemented and (b) that didn’t change anything in the 2.0 spec. We had thought it would take a year at most. There was long debate about putting all of our effort on VoiceXML 3.0 - which was planned to be a large change in the core spec - instead of spending time on 2.1.

We did the original spec authoring relatively quickly and published the first working draft in the spring of 2004. At the time, I was one of the primary co-editors of the spec along with Ken Rehor, our chief architect at Vocalocity and the original author of VoiceXML 1.0 and 2.0 spec. Like Tellme, Voxeo, Voicegenie, Voxpilot and Loquendo (and probably others), we had all agreed between us on the core spec features and had all implemented the spec when we authored it. So - this should have been a slam dunk. Then, enter the W3C process.

The W3C process is generally a good process. However, like any process, it has it flaws. The process is much different from the IETF process of “rough consensus”. They both represent two extremes in how a standards process can work. While the W3C takes almost herculean strides to make everyone have a say (and present a challenge which stops progress), the IETF looks for general consensus and never counts votes. IETF meetings are famous for “hum votes” - instead of raising your hand if you agree with something, in IETF meetings, they might ask for a “hum” or a “finger snap” vote. The louder the noise, the more positive (or negative) the vote. Also, in W3C meetings, in person voting and meetings are largely more important and final than the e-mail list. In the IETF, you can never attend a meeting in person and have plenty of authority and input - in fact, any in person vote has to be confirmed on the list.

I won’t go into all my personal reasons why I believe that the VoiceXML 2.1 spec took so long. We had a number of ridiculous challenges from some people outside of the Voice Browser community about the specification - people who had no vested interest in the success or failure of the spec and would never implement it - but wanted to get their input taken. We had a number of “fights” about how to deal with procedural issues because of the W3C beaurecracy and a number of procedures we had to follow that really didn’t apply to the group or the spec itself. In the end, I’m happy to see after 4 years of development on a spec that took us only a couple of months to author - it’s finalized. What’s more funny is how the industry and the people involved have completely changed since we first started out:

Matt Oshry from TellMe — Matt was the chief cat herder of the 2.1 effort and TellMe was acquired by Microsoft earlier this year. Funny enough, when I started working with the Voice Browser working group - Microsoft was very very active and vocal. They then one day stopped caring about the spec. And then shortly afterwards, introduced a competing specification called SALT. SALT failed very publicly less than a year later and Microsoft later came to my company (Vocalocity) and decided to OEM our browser for their telecommunications product. And now, with the TellMe acquisition, their one of the largest VoiceXML vendors in the world. Michael Bodelle was also from Tellme and very involve early on in the development.

Brad Porter from Tellme — Brad was also one of the key architects of the spec. Brad is now with Amazon running their web services platform.

Paolo Baggia from Loquendo — Paolo is still with Loquendo and probably one of the only remaining independent companies that has survived.

Dave Burke from Voxpilot — Dave is now working on mobile for Google and relocated from Dublin to London.

Dan Burnett and Jerry Carter — Both Dan and Jerry joined us after leaving Nuance for Vocalocity. Dan left after the acquisition and Jerry just before and went back to Nuance to continue working on standards. Dan was also the chief editor for the SSML specification.

Scott McGlashan of HP - Scott’s company Pipebeach was acquired before the 2.1 spec was initial released by HP and he’s been working with HP since.

Alex Lee and Mark Scott from Voicegenie - Voicegenie was acquired about a year or so ago by Genesys.

Emily Candell from Comverse - Comverse is a large public company and there still doing fine and I assume Emily is still with them.

RJ Auburn - RJ is still the CTO of Voxeo and Voxeo is doing well. A number of people from Vocalocity went to work for Voxeo after the acquisition.

Ken Rehor - after Vocalocity, Ken has been continuing to be a leader and evangelist in the VoiceXML community. Enough great things can’t be said about Ken and his efforts.

Congratulations to the voice community for getting this spec through - it’s been a long time coming.

technorati tags: w3c, ietf, voicexml, ssml