RE: ZDNet Schema article,and hiding complexity withinuser-friendlyproducts

[Gavin Thomas Nicol <gtn@ebt.com>]

> Fair enough.  Nevertheless, hardly anyone has actually USED the
> beast yet except for the kinds of people on this list. 

I think a more telling point is that people don't use validation
that much anyway... be it using DTD's, XSchema, or whatever. 
What percentage of people on this list *really* use validation
as part of normal processing?

Much of this debate might be a storm in a teacup.

XML's success is at least partly due to it making things easier
for people... including those people with SGML systems. We should
remember that. This is what Len is reminding us of.

The real danger, I think, is *not* in the specification. XSchema
is large, dense, and probably overkill for most applications. For
the people that need it though, it is probably both approachable and
very useful.

The real danger is in assuming, and having specifications assume,
that validation will be part of normal processing. For example, 
if XPath 2.0 is expanded to include operations that require a 
validated instance, there will be interoperability issues: if
I send an XML instance and an XSL stylesheet to different 
processors, the might produce different results. XPath is already
very close to being too heavyweight.

A good example of the issues can already be seen in the differences
in documents parsed with and without DTD's. If validation with
XSchema or any other tool results in an infoset with many "optional"
peices, and specs are built on that infoset, there *will* be
issues.