form/menu grammars with document scope?

Am I reading the spec (VoiceXML 2.0) right? It seems that after
loading the document (for example using Xerces Java library), I have
to scan all top level dialogs (form/menu), go inside them and look
into each grammar element (possibly within link element?), check if
the scope is document. If so, then install and activate the grammar
for the document.

And further, when entering a dialog element, I need to scan the child
grammars again, and this time only install (activate) those that
didn't specify document scope.

I suppose this is implementation specific, but I was trying to "object
oriented" solution, where each element only implements "it's own
thing" with lightweight interpreter engine that just "runs" the
elements.

There are some other VXML elements that have to peek at their parents
and do some digging into child elements. I can sort of accept single
level up (parent) and down (children) "snooping", but the grammar
scope thing needs much more than that, and I'm somewhat unhappy about
it...

I don't really have anything good to propose, but an idea

- scrap the "scope" attribute. Specify scope is always from the
  context where the grammar appears,

- allow grammar elements under <vxml>. These will be document scope.

-- 
Markku Savela

Received on Friday, 30 November 2001 11:35:44 UTC