I wrote:
> Elliotte Rusty Harold wrote:
> > It certainly doesn't have to validate but a conforming XSLT processor
> > *must* read the external DTD subset. The reason is that the XPath
> > data model does not allow unresolved entity references. If any of
> > these are defined in the external subset, then you can't build a data
> > model for the document without reading the external subset.
>
> No, this isn't right at all. Skipped entity references are simply
> invisible
> to the data model, i.e. lost, just like anything else that's not
> significant
> in the data model. This is a very important distinction.
> Otherwise, it would
> be impossible to use a non-validating parser with XSLT.
More precisely, it would be impossible to use a non-validating parser that
meets the minimal conformance requirements for XML parsing.
Evan
