> I was just arguing against the thesis that 
> variety is good; in my view it is a necessary 
> evil caused by the inadequacies of the currently
> available languages, starting with DTDs.

Actually, I don't think that DTDs are inadequate at all.  DTDs fill a
very useful niche.  If all I'm interested in is to define an XML
vocabulary and am not concerned about data type checking (perhaps
because my database is already performing that role), then DTDs are
perfectly fine.  Apparently, other people share this view (of the
adequacy of DTDs); for example, XHTML documents are validated against a
DTD.

Likewise, if I am interested in expressing co-constraints of data
within and across XML documents then all I need is Schematron.  I
certainly don't want to be burdened with the overhead of a
grammar-based language.  Thus, Schematron serves a very useful niche.

"Use the right tool for the right job" is an old adage that seems to be
pretty fundamental.  Just as one programming language is better suited
to a task than another, so too, one schema language is better suited to
a task than another.  Just as variety is rampant and useful in Nature,
variety is rampant and useful in programming languages, schema
languages, and the Web itself!

To summarize, each of the currently available schema languages fill a
niche.  What's needed now is the ability to use them collectively.
More importantly, however, what's needed is for the XML community to
rise to a new abstract level of thought; we need to move away from the
mentality of "using a monolithic schema" to the mentality of "using XML
vocabularies" (wherever they may hail from).  And that's exactly what
NVDL provides.

/Roger