> The current grammar-level inclusion proposal allows us to write:
>
> <externalRef href="XHTML.rng">
> <start>
> <ref name="div"/>
> </start>
> </externalRef>
It is not intended to. Only grammar-level include's can contain <start>,
and <externalRef> was propsed as an alternative name for the pattern-level
include. What the proposal we adopted allows is:
<grammar>
<include href="XHTML.rng">
<start>
<ref name="div"/>
</start>
</include>
</grammar>
Is that what you meant?
> I believe this kind of simple references is quite common.
> For this purpose, I'd like to propose an optional attribute "name" for
> the <externalRef> element.
>
>
> By using this attribute, the above pattern can be simply written as:
>
> <externalRef href="XHTML.rng" name="div"/>
>
>
> Also, this makes <externalRef> look closer to <ref>.
At the moment externalRef (pattern-level include) doesn't care what kind of
pattern you reference; it might be a <grammar> pattern, or it might not.
When externalRef has a name attribute, then it would have to be referencing
a <grammar> pattern.
One advantage of this proposal is an implementation one. With externalRef,
it's easy for an implementation to maintain a table mapping URIs to
patterns, so that it doesn't have to reread a a file every time it is
externalRef'ed. Adding a name attribute wouldn't break this. However, for
a grammar-level include, an implementation will typically need to reprocess
the include every time it is encountered, because the included grammar gets
merged with the including grammar.
James
