Re: ANNOUNCEMENT: RDFStyles: alternative to XSLT for RDF

On Wed, 22 Oct 2003 09:36:41 -0400, "DuCharme, Bob (LNG-CHO)" <bob.ducharme@lexisnexis.com> wrote:

> Graham Klyne  wrote:
> >At 11:37 21/10/03 -0400, DuCharme, Bob (LNG-CHO) wrote:
> >>These syntactic variations are RDF's problem, not XSLT's, any more than
> >>rendering of SVG or XForms are XSLT's problem. In other words, the root
> >>cause of the sadness you describe is at
> >>http://www.w3.org/TR/rdf-syntax-grammar/, not at
> http://www.w3.org/TR/xslt.
> 
> >I find comments like this to be unhelpful.

Me also, and that's why until now I've been ignoring this thread; which needs retitling.

> I did make one mistake there which apparently gave the wrong impression:
> instead of saying "These syntactic variations are RDF's problem," I should
> have said "RDF/XML's problem." (After all, RDF itself doesn't have a syntax,
> right?) It looked to me like people were blaming XSLT for XML/RDF's
> problems, so I pointed to the XSLT spec to make it clear that recognizing
> RDF triples is not part of XSLT's job, and I probably should have pointed at
> http://ilrt.org/people/cmdjb/2003/05/iswc/paper.html#sec-reported-probs
> instead of the RDF/XML spec to more clearly make my point that the problems
> that typical XML processing tools have with processing RDF are not the
> tools' problems, but RDF/XML's. 

That paper by me (I'm the editor of RDF/XML's revision) just lists some
reported problems over the revising period.  They weren't all big
problems and several have been fixed since then - it's a paper reviewing
the history.

> >XSLT is a pretty good hammer. But why does anyone think that XSLT is the 
> >right general-purpose tool for manipulating RDF information?  Beats me.
>
> It beats you because you're abstracting this up to the level of "RDF
> information." No one considers XSLT to be the right general-purpose tool for
> manipulating RDF information. XSLT is for processing XML.  ...

XSLT is for processing XML - hierarchies of tags and content.  You write
some specific XML<>XML transformation from some XML format to some other
format.  (You can't use XML schemas here, lets stick to XSLT 1.0). That
XSLT will be specific to the document structure of both the source and
destination, as well as the tags used in both of them.

This is very similar to applying XSLT to RDF/XML describing some
vocabulary.  True, you likely will have to live with working on some
RDF/XML syntax profile - you'd have to, since you have to know some of
what vocab to expect. This approach been used successfully several times.
This is probably what you are likely to have to do with any XML, given the
variety of XML formats available.

> ...The W3C gave the
> world an XML expression of RDF. People want to use their XML manipulation
> tools to process this XML, and they're frustrated that they can't. Some
> blame the tools themselves, which is what I was objecting to. 

I bet you could have similar fun or disasters with SVG or XML Topic Maps.

If I dumped tables in my database into XML in some set of tags and then
used XSLT, I could hardly complain when the resulting XML wasn't a
proper database representation.  Or if I took some XML and applied
text-processing tools that ignored tags, I could hardly complain when
the XML wouldn't validate.

XSLT is sometimes appropriate for XML problems and it is sometimes
appropriate for RDF/XML problems also - both require taking care.

> ... IIf the problem
> can be fixed at the root, which again is the existing RDF/XML expression of
> RDF and not RDF itself, we'll all be better off, which is why I was happy to
> see Tim Bray's http://www.textuality.com/xml/RPV.html and look forward to
> further proposals for alternatives.

RPM was one step and incomplete as I discussed in
  http://ilrt.org/people/cmdjb/2003/05/iswc/paper.html

The goals for new syntax(es) for RDF remain confused and mixed
between end user syntaxes (RPV, others) and canonical/normalized
syntaxes (typically called XML N-Triples).

> I'll bet a normalized representation of triples in XML plus some of the
> extra features of XSLT 2.0 would make it possible to write some genuinely
> useful RDF applications with XSLT. I will look into it. 

In any new syntax, it will remain that people will still see a hierarchy
of XML elements when the model is a set of triples, and likely make similar
mistakes.  It's possible a better syntax would help reduce that, but you
can't legislate that! :)

Dave

Received on Wednesday, 22 October 2003 11:45:45 UTC