Neil Graham wrote:
I think my question revolved around Joerg's assertion that the spec thinks
of XInclude processing being done before validation.
Sorry, I wrote
>>> The REC itself is deliberately neutral.
XInclude is defined in terms of "merging infosets", and it is briefly
and non-normatively) discussed what this could mean if validation
occurred before or afterwards.
I personally see there are use cases for either case:
1. XInclude as a substitute for external entities for physically
structuring documents. This is useful if the document should
be validated against a schema, where entities are not organically
available. Naturally, the post-XInclude infoset has to be
validated.
2. Validation before XInclude. This could in turn mean the including
document is validated, or the included documents, all documents
or arbitrary combinations. This use case happens for formats which
are meant to aggregate content, for example imagine a portal where
you aggregate vastly different content, like an article, stock
quotes and ad streams. While each of the aggregated content may
have an attached schema and therefore may be easy to vgalidate,
the aggregated content may prove more diffucult in this respect
(imagine if they use different schema languages).
Therefore parsers with an integrated XInclude stage (before validation)
are useful (remember libxml), but standalone XInclude processors make
sense as well.
J.Pietschmann
