RE: Question/comments about cross-document ID/IDREF for XInclude

[Reposting to comments list, as WGs often have need to refer to these types
of conversations later, hope that's all right with you.]

Unfortuneately, when validation is performed is a murky area.  DTD
validation is by definition a parse-time process, and there is no
specification which describes how to DTD-validate an infoset.  So we would
have a lot of work to do that is not central to our charter to describe how
DTD validation on an infoset would be performed.

Likewise, XML Schema does not describe exactly when validation occurs, but
only how to validate an infoset.  If Schema doesn't even describe when
validation occurs, why should we?

The WG decided that we would follow the lead of XML Schemas and describe an
infoset-to-infoset transformation, without limiting the scenarios in which
it may be applied.  The current draft doesn't prevent you from the scenario
that you seem to want, which is to parse a document with a
non-DTD-validating parser, process the infoset using an XInclude processor,
and validate the result using an XML Schema processor.  In fact, this may be
the most common useage of XInclude, and one that I expect vendors to support
in an integrated way.  Do we need to do more in the spec to support this
type of product?  Can we actually do so in a way that doesn't limit other
scenarios?


> -----Original Message-----
> From: Morten M. Christensen [mailto:mmc@mortench.net]
> Sent: Tuesday, December 12, 2000 5:38 AM
> To: Jonathan Marsh
> Subject: Re: Question/comments about cross-document ID/IDREF for
> XInclude
> 
> 
> Hi Jonathan Marsh,
> 
> Thanks for your reply! I am glad that XInclude support the 
> scenario I was
> thinking about regarding multiple-docs, ID, IDREF and validation.
> 
> An yes, ID/IDREF (validation) appears to be orthogonal but at 
> least the XInclude
> spec should describe in which way... For example it could 
> state that conforming
> XInclude implementations MUST be able to perform validation 
> AFTER inclusion, to
> support documents that are invalid in parts but valid as a 
> whole. - And that
> specifically covers validation issues like ID, IDREF across 
> document parts.
> 
> --- Morten Christensen
> 
> Jonathan Marsh wrote:
> 
> > We had an issue on this in a previous draft - the 
> resolution was that IDs
> > are merely a result of schema validation and thus was 
> orthogonal to the
> > XInclude draft.  The validation method determines what you 
> might expect from
> > the result.
> >
> > For instance, if IDs are declared via DTDs, the symbol 
> spaces of the two
> > documents is combined, and IDs should work as if they were 
> constructed as a
> > single document.  If there are duplicate IDs, you may get a 
> document that is
> > valid in parts, but invalid as a whole - and this is not 
> limited to just
> > duplicate IDs.
> >
> > With Schema validation performed after inclusion, both the 
> combined content
> > model and the uniqueness of IDs can be checked.  I think 
> this is likely to
> > be the more common scenario.
> >
> > > -----Original Message-----
> > > From: Morten M. Christensen [mailto:mmc@mortench.net]
> > > Sent: Friday, December 08, 2000 12:12 PM
> > > To: www-xml-xinclude-comments@w3.org
> > > Subject: Question/comments about cross-document ID/IDREF 
> for XInclude
> > >
> > >
> > > Hi,
> > >
> > > I have looked at the working draft of XML Inclusions
> > > (XInclude) Version
> > > 1.0. While I support an include mechanism, I am mystified that the
> > > proposal does not mention what I see as the the main 
> issue of ID/IDREF
> > > across the including and included documents (ie. an 
> included document
> > > refer with an IDREF to the ID in the main document that 
> contain the
> > > include tag).
> > >
> > > I would like to ask what/how XInclude will solve that problem?
> > > Personally, I would be much relived if XInclude does 
> support ID/IDREF
> > > across  including and included documents...I know it is 
> probably hard
> > > technically, but it would solve a lot of problems for may 
> people and
> > > eliminate many difficulties in using XInclude!
> > >
> > > Sincerely,
> > > Morten Christensen
> > >
> 
>

Received on Wednesday, 13 December 2000 13:12:15 UTC