RE: XML or RDF Re: The Tragedy of RSS

>>For example, is the following XML document an example of the use of
>>RDF? (ie. it is RDF?)
>>
>><?xml version="1.0" encoding="UTF-8"?>
>><foo xmlns:dc="http://purl.org/dc/elements/1.1/">
>>  <bar>lorem ipsum</bar>
>>  <dc:title>This is a title</dc:title>
>></foo>
>>
>
>This question ties to a few others that have come up.
>
>My response would be this is just XML, since nothing says otherwise.
>
>The caveat is that if we allow RDF to omit rdf:RDF and rely on well-known
>namespaces then we have to figure out a way to answer the question for a
>parser - i.e. is it legitimate to rewrite this as
>
>  <rdf:RDF>
>    <rdf:Description about="">
>      <dc:title>This is a title</dc:title>
>    </rdf:Description>
><foo id="barfoo">
>  <bar>lorem ipsum</bar>
></foo>
>  </rdf:RDF>
>
>or not?

I'd say it would depend on the model & syntax used by foo. DC title would
map across directly whatever.

In another post Mike Daconta mentions using DC in RSS 2.0 in this fashion -
that makes a lot of sense. If Dave Winer can get someone to write the
namespaces part of his spec, then RSS 2.0 will overnight gain a huge amount
of power. Tidying up the use of GUIDs/URIs would be another step forward.
The model would still remain somewhat underdefined, though if this was
worked on as well then RSS 2.0 could have the same expressive power as RSS
1.0...

>We could also wonder what the namespace says - does it provide an
>XML schema,
>or an RDF schema, or something else, or both? Does it answer to a
>request for
>something that transforms from one to the other?

Being pragmatic, I'd say 'no' to all of these. RDDL was a nice idea to
manage this kind of thing, but it doesn't seem to be finding much adoption.
If the namespace is simply viewed as identifying a virtual dictionary, and
the terms are defined *somewhere* (in any language), then this isn't a
problem. Transformations are in the province of the languages in question,
and behaviour will be local, i.e. application specific. All we know, and all
we *need* to know is that when I talk about dc:title is that it means the
same as when you talk of dc:title.

Ok, in an ideal world maybe all terms would be formalised, given generally
applicable mathematical definitions and syntactical constraints using
schema, but for most purposes I reckon the need for this falls outside of
the current 80:20 situation. A human readable doc saying xyz:name is a
predicate which should be a unicode character string is adequate. Where the
definition really is required (DAML+OIL or OWL or whatever) then the extra
semantics are a bit too meta-language for XML or RDF Schema, so the mapping
through namespace probably wouldn't work anyway.

Cheers,
Danny.

Received on Saturday, 5 October 2002 06:34:01 UTC