[
Lists Home |
Date Index |
Thread Index
]
On Jun 4, 2004, at 10:29 AM, Elliotte Rusty Harold wrote:
>
> Simply put, I don't care if adding additional information breaks some
> schema someplace. Properly designed XML applications ask whether a
> document contains the information need. They do not ask whether the
> document contains additional information they don't need or care
> about. I agree that XML document should be Extensible. I disagree
> that plain vanilla XML documents aren't extensible. Validity is not
> required.
Yes! One popular use case for RDF seems to be as sortof a lightweight
database for unstructured data. That's fine, and if people can
leverage the triples model with other tools, or really believe that the
logical network implied by the data is consistent enough to do logical
inferencing, so much the better. BUT simple well-formed XML + XPath,
XSLT, and/or (basic) XQuery can handle what Mark is talking about just
as well, and there are probably a lot more tools out there that support
XML+XPath than XML+RDF.
Mark's comment that the triples datamodel ' presents information in
discrete "packets" ' intrigues me, however. As he is fond of pointing
out, architectural constraints are a Good Thing in many respects even
though they limit options. I personally suspect that XML elements can
be nice "discrete packets" that hit a better compromise between
constraint and usability than RDF triples, but the point is well taken.
As I understand the evolution of RDF thinking, the current idea is to
think of triples as the "assembly language" rather than the surface
syntax. What optimism about the semantic web technologies that I now
have comes largely from the hope that the power of XML+Xpath and the
power of logical inference enabled by the triples/SPO model can be used
in harmony, rather than as alternative paradigms.
|
