That is debatable...and has been debated. The DTD syntax would indicate
that tehre is no relationship, but the fact that under the namespace
proposal they can be namespace qualified suggests that attributes can be
shared. Most people who read, for instance, the XLink specification is going
to believe strongly that those attributes are shared. The same goes for
the xml:lang attribute. It stretches intuition to call them merely
different attributes with similar names.
> > If we want to allow nesting of attribute constraint declarations in
> > element constraint declarations, it should be a short-hand for the
> > expanded version.
>
> I disagree. I am willing to accept a shorthand, either by way
> of general entities in XSchemas or otherwise, for copying
> attribute declarations from one element to another by way of
> shorthand, but not for reifying attributes independently of the
> elements they are attached to.
General entities are an ackward mechanism for reuse, and have essentially
all of the flaws of XML parameter entities. One of our goals should be to
reduce the need for them. I can't think of another syntax that
> Perhaps the practical result would be the same, but the theoretical
> implications are very different.
The SGML standard allows attribute declarations to be shared, but does not
reify them. It is all a matter of whether you interpret the syntax as
shorthand for actual sharing. Personally, I am in favour of the
(IMO) intuitive reification interpretation, but I don't think that we
need to make that choice ourselves. We are defining a language for
constraining XML documents, not for driving the interpretation of XML
documents. Applications and authors can interpret attributes with the
same name however they want...we just happen to provide a nice notation
for sharing declarations of constraints.
Paul Prescod