Just to clarify from before, this is how the VXML infrastructure works:
[Removed: ASCII formatting didn't turn out. See attached file for diagram.]
You host your VXML pages on any Internet webserver (e.g. chiefdelphi.com, yahoo.com, geocities.com, &c.). Then, you find a VoiceXML Gateway that has something called a VoiceXML Browser. Gateways include: TellMe Studios, VoiceGenie, &c. Finally, you get a phone number (or an extension) that you can call to reach your VXML page
via the VoiceXML Gateway.
All the speech recognition and synthesis is handled by the VoiceXML browser. The webserver serves up the VXML file to the browser, which in turn, reads it to you on a telephone.
Like you said, really cool stuff. *grin* It was just starting to gain steam last year, and it's becoming more and more popular. Soon, there may be a large influx of VXML developers, and free VoiceXML gateways will pop up like the GeoCities of today.
By the way, anyone interested in helping out, we need some voice talent. Someone to read off an introduction and give various canned lines. If you're interested, email me:
cohesion@mediaone.net.