xml-dev - Best Content Model Re: XML Schemas: Best Practices

[Rick JELLIFFE <ricko@geotempo.com>]

Can I make a comment that the best practises documents must fight hard
to stop the tendency to use as the null hypothesis data element 
 <element name="Title" type="string"/>
rather than
 <element name="Title" type="anyType"/>
or even the equivalent
 <element name="Title" />
which (I think) is or will probably be available.

The reason is that it is an internationalization requirement for text in
Chinese and Japanese to support Ruby annotations and for text in Arabic,
Hebrew and perhaps Chinese to support BIDI (bidirectionality) elements.

So, for general text, only make an element simpleType string if there is
a hard requirement that it does not contain free text, that it is not
going to be used for the languages above, that it will not be annotated
or have other elements inserted.

Instead of using a type derived from String, use a type derived from
anyType with mixed content and wildcard it to allow subelements from
other namespaces only.  Allow xml:lang.  That should be the typical way
to declare anything that is likely to be free text.  Otherwise, as soon
as someone wants to annotate or use has international requirements, they
have to break the typing somehow. 

So, in general, String should be used for constraining formal data only.
The best practise for general text is a slightly open (i.e., data
allowed and other namespace elements) content model.

Cheers
Rick Jelliffe