The question in my mind is "will the
industry REALLY accept it as a Standard?" I have no idea ... We as a community
need to figure out:
- Can the spec can be implemented by those not
closely involved in its creation? (for example, there are significant bits in
the first incarnation of XML 1.0 that assumed a knowledge of SGML; will the
ability to implement the XSD spec depend on shared knowledge that it not written
down in the spec itself?)
- Are implementations interoperable in
practice? We need something like the OASIS XML test suite, and someone to
do for XSD what David Brownell did for XML to rigorously analyze the degree to
which different implementations of the spec yield the same results from the same
inputs.
- Where is the "sweet spot" for ordinary users? In other
words, what is the "Common XSD" subset that ordinary users can easily understand
AND all the implementations handle properly?
- What's the real business
case for selling XSD to non-geeks? XSD MAY add more value to
real business systems than it costs. There is a theoretical case to be made for
why a declarative constraint specification language such as XSD is "better" than
procedural code that validates business rules ... but can we REALLY use XSD to
build practical examples that prove it?
- If XSD can't meet some or all
of these criteria, can TREX and/or RELAX do a better job? And if they do,
will the CTO's of the world care?