RE: [xml-dev] Validate **against** a schema OR validate **with** a

["Michael Kay" <mike@saxonica.com>]

> > Defaulting values came from SGML days and DTDs..  So it's 
> been around 
> > for a while.  Many folks have wanted the validation and 
> augmentation 
> > steps separate though no real progress on that.
> 
> Adding default information and other properties is a very 
> respectable thing for a schema language to do, IMHO.
> 

I think that the database community has actually done a better job in this
area than the SGML/XML community.

In a database schema, you typically distinguish:

(a) structural constraints (which columns go in which tables, the equivalent
of our grammars)

(b) integrity constraints (the equivalent of our assertions, as in
Schematron or XSD 1.1)

(c) view definitions (subsetted or augmented or otherwise transformed
information that can be exposed to an application)

A key point about view definitions (or subschemas, as they were called
before the relationalists took over) is that you can have many of them, so
different applications can see different views. We don't seem to have
grasped that idea, though we struggle with the consequences. And subsetting
is at least as important as augmenting. Another point which SQL got right is
that you can use the same language to define queries as to define views, so
once you have a working query you can treat its result as a view. This of
course demands closure, that is, the output of a query/view uses the
original data model, which relates back to the point Rick made about the
PSVI being expressible in XML.

(Once again, Roger has triggered an interesting discussion by asking what
appeared to be a really trivial question!)

Michael Kay
http://www.saxonica.com/