RE: [xml-dev] Will the next version of XML Schema have aschema-for-schem

[Rick Jelliffe <rjelliffe@allette.com.au>]

On Wed, 2007-10-17 at 16:07 +0100, Michael Kay wrote:
> > As I understand it, the next version of XML Schemas will 
> > provide a way to express co-constraints.  So, in the next 
> > version of XML Schemas will the schema-for-schemas be 
> > expressible entirely using the XML Schema language? 
> 
> No, there will still be some rules for schema validity that can't be
> conveniently expressed within the language, for example the rules governing
> the syntax of XPath expressions.
 
And, in fact, it can a bad thing to require a schema language to be
completely specifiable in terms of itself.  

In XSD, for example, the use of elements to describe complexContent and
simpleContent is jarringly odd and unpleasant, and it comes from the
fact that XSD cannot use attributes to select a content model. Elements
is all it has in this situation.

Therefore, in XSD anything that impacts the content *must* be described
using (ultimately) the element name, not an attribute. (@xsi:type is a
special case.) So when using XSD, you cannot use attributes to be
attributes that specialize the generic identifier, you have to use very
specific identifiers. It is no advance on 1986 DTDs.

This is because XSD really thinks of attributes as being funny
sub-elements, rather than the element name being a funny truncated kind
of attribute (which is the "architectural" view). And it is because
there is no way for the children of an element to determine the
effective type of the parent (err, except for the special and partial
case of derivation by union.)

Cheers
Rick Jelliffe