Jim Larson, co-chair of the W3C Voice Browser Working Group, recently posted on the www-voice mailing list that work is well underway on VoiceXML 3. He wrote:
The purpose of V3 is to provide powerful dialog capabilities that can be used to build advanced speech applications, and to provide these capabilities in a form that can be easily and cleanly integrated with other W3C markup languages.
That’s very good to hear. The SALT specification clearly provides better integration with HTML. It’s good to see that some of the better aspects of SALT will make it into an official standard.
Larson also wrote:
V3 will offer platform vendors the choice between VoiceXML 2.x or dialog markup languages that they define themselves. Cross platform support is ensured through an interoperable means to bind markup components to definitions built upon a set of primitive objects that support speech recognition, prompts, DTMF, telephony, voice verification, recording and playback. Below the object layer is an implementation layer consisting of platform functions, capabilities and technologies available either locally to the platform itself or remotely via access to services elsewhere.
I think I’ll need to know a bit more about this proposal before passing judgment on it.
One rarely advertised aspect of VoiceXML development tools and platforms is that the tools rarely generate portable, static VXML. In almost every case, the tools generate some combination of static XML and/or properties files along with servlets, JSPs, and/or Java classes that dynamically generate the VoiceXML. So, even though the end result is relatively standard VXML mark-up, the VXML is typically not generated until run time. As a result, the applications you develop with these tools really aren’t that portable. There are a few exceptions to this rule, but most of the exceptions are tools that are useful for only the most trivial of speech apps.