On Wed, 13 Mar 2002, Simon Cox wrote:
...
> Yeah - I think this might be right.
> Note that it is not /XML/ that is deficient here, it is W3C XML Schema.
> The main advantage of XML Schema is its data typing system.
I've growing strong affections for relax-ng [1] (and I'm also very fond
DTDs). It comes relax-ng does not have the notion of inheritance as in
XML Schema, but it can be modularized [2] in a more transparent ways than
DTDs. And it has support for multiple data-type libraries, and there
is a brief discussion on how to use XML Schema Datatyping with
relax-ng[3].
I think, for example, that the modularity of relax-ng would be sufficient
for satisfying Carl's wish for only supporting dc with mathml.
> However, that comes at a cost.
> One of the limitations is that "simple" types are in a completely different
> hierarchy to "complex" types (i.e. ones with sub-elements).
> As I noted, the hierarchies only join at one point - the top.
A relax-ng grammer can provide a default pattern for a XML construct, but
may also load extensions with more detailed patterns. For instance modules
can be added that could provide support for subject encoding schemes.
> Then there are some tricks that can be pulled with "mixed" types - i.e.
> element content which includes text interspersed with elements.
> This all means that it is very tricky to build a base system which it is
> possible to refine to both simple and complex types.
> And this is at the centre of the simple vs qualified DC issue.
Sigge
[1] http://www.oasis-open.org/committees/relax-ng/spec-20011203.html
[2] http://www.oasis-open.org/committees/relax-ng/tutorial.html#IDA04YR
[3] http://www.oasis-open.org/committees/relax-ng/xsd-20010907.html
|