RE: [xml-dev] The spirit of XML (was RE: [xml-dev] Re: I can XInclude where I [expletive deleted] wantto)

["Evan Lenz" <evan@evanlenz.net>]

Uche Ogbuji wrote:
> It actually wouldn't be too hard to do: 4Suite uses expat for
> non-validating parse, and all we'd have to do is provide a flag
> which causes ExternalEntityParserCreate to be invoked rather than
> ParserCreate (or whatever the precise function names are).  IIRC,
> we still need to set up a fake top-level parser to satisfy expat,
> but shouldn't be much sweat.
>
> Your post might be the prodding I need to finally do it.

Cool :)

> Of course, it's not just a matter of parsers, eh?  I know the
> basis of TransQuery is that one can have the XML database look
> sort of like an oversized EPE whereas each top-level node
> represents a separate document in the database, right?

That was one idea, but things like IDs and base URIs that require document
identity to be maintained forced me to stick with viewing the XML database
as a simple node-set containing only root nodes. Access to that node-set can
be had through the transquery:input parameter. That's TransQuery in a
nutshell. Individual documents, of course, can be EPEs (e.g. boilerplate
text, document fragments, etc.), and should still be individually accessible
via the document() function. Ultimately, the need to do joins between
different documents will be much aided by XPath 2.0's "for" expressions,
whether in the context of XSLT 2.0 or XQuery. XSLT 2.0 (and certainly
XQuery) may deal more explicitly with document collections, thereby
rendering TransQuery irrelevant, except in an XSLT 1.0-only context. One of
the main goals of the project was just to draw attention to the idea of
implementing XSLT over an XML database, something that I would still like to
see happen some day.

Evan