Re: FW: how does XInclude mix with XML Schema? XSLT?

On Wed, 2004-09-08 at 13:03, Norman Walsh wrote:
> Dan,
> 
> I don't believe there are any tests of XInclude used with XML Schema
> or XSLT processors. I don't know of any processors that (directly)
> support XInclude in schema documents or stylesheet documents.
> (Indirect support through pipeline languages such as sxpipe is
> certainly possible.)
> 
> Schemas can certainly be written that validate the occurence of
> xi:include elements in input documents (assuming that the input
> document was not subject to XInclude processing, of course).
> Similarly, stylesheets can be written that will transform xi:include
> elements or produce them in the transformed result.

Would you (plural) please write some and put them in a test suite?

> 
> An XInclude in a schema or stylesheet document, if the document was
> subject to XInclude processing before the schema or stylesheet
> processor encountered it, would behave exactly like an external parsed
> entity.
> 
>   <xsl:stylesheet xmlns:xsl="...">
>     <xsl:output method="xml"/>
>     <xi:include xmlns:xi="..." href="someDocument.xsl"/>
>   </xsl:stylesheet>
> 
> is effectively the same as
> 
>   <!DOCTYPE xsl:stylesheet [
>   <!ENTITY someDocument SYSTEM "someDocument.xsl">
>   ]>
>   <xsl:stylesheet xmlns:xsl="...">
>     <xsl:output method="xml"/>
>     &someDocument;
>   </xsl:stylesheet>


That seems worth capturing in a test.


> This is distinct from the xsl:include element (or the xsd:include
> element) which performs additional, application-specific semantic
> processing *in addition to* performing a kind of textual inclusion.
> 
> Had XInclude existed when XML Schema or XSLT were under development, I
> think it *could* have been used instead, but it cannot be used instead
> now.

That's another assertion that seems to merit a test.

>  In particular:
> 
>   - including one stylesheet "style.xsl" in another with XInclude
>     would likely result in a xsl:stylesheet element as a child of the
>     "including" xsl:stylesheet which would be an error.
> 
>   - including one stylesheet "style.xsl/*/*" in another with XInclude
>     would not inform the including stylesheet of any attribute values
>     that might have been set on the included stylesheet's root element.
>     That could produce quite different semantics than xsl:include.
> 
> I hope this helps clarify the situation.

Well, yes, but I'm afraid it makes it clear that the situation
is not, yet, as it should be.

>                                         Be seeing you,
>                                           norm
-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/