Re: ConceptSchemes and DC metadata

Norman,

On Fri, Jun 19, 2009 at 06:40:03PM +0100, Norman Gray wrote:
> The SKOS Primer illustrates annotating ConceptScheme instances using  
> DCTerms properties.  Is it a recommendation to use that rather than  
> the DC elements properties from <http://purl.org/dc/elements/1.1/>?

Yes.  This point was raised and discussed in the SWD working
group (e.g., [1]).  As I sometimes put it, DCMI "gently
promotes" the use of the dct: equivalents of the fifteen
elements because of their formal ranges.

> This is possibly more of a DC question than a SKOS one, but this list  
> is probably as good a place as any to ask the question, especially  
> since I'm asking it in the context of trying to describe 'good  
> practice' for developing vocabularies.
> 
> DC elements have the advantage that they're probably more generally  
> understood, and more things might be declared as rdfs:subClassOf DC  
> elements.
> 
> On the other hand, the DCTerms properties are more expressive, are  
> declared as subclasses of the DC elements properties, and are newer  
> (though I couldn't find an explicit statement that the older ones are  
> deprecated, beyond the incidental description of the DC Elements  
> properties as 'legacy' in the DC Terms documentation).
> 
> My guess would be that the DC Terms properties are what I should  
> recommend -- is there anything wrong with that?

DCMI has gotten alot of positive feedback from the Semantic Web
community on the "makeover" of DCMI properties with domains and
ranges.  A longer explanation, with historical context:

dc:title [1] and dc:subject [2] (and the other thirteen Dublin
Core properties) were among the first RDF properties declared
anywhere.  They were declared as RDF properties before W3C
standardized the notion of "range" in the RDF Schema
specification.

As RDF matured, the DC properties became criticized in SW
circles for being underspecified.  DCMI wanted to assign ranges,
but in doing so did not want to "break" existing legacy data,
which used "subject" (for example) both with literal and
non-literal values.

As described in [4, paragraphs starting "Formal domains..."],
DCMI resolved this dilemma by creating fifteen properties in the
/terms/ namespace in parallel to the corresponding terms in the
/elements/1.1/ namespace, and declared the former as
subproperties of the latter.

It is not actually incorrect to continue using dc:subject and
dc:title -- alot of Semantic Web data still does -- and since
the range of those properties is unspecified, it is not actually
incorrect to use (for example) dc:subject with a _literal_ value
or dc:title with a _non-literal_ value.  However, good Semantic
Web practice is to use properties consistently in accordance
with formal ranges, so implementers are encouraged to use the
more precisely defined dcterms: properties.

Tom (wearing his DCMI hat)

[1] http://dublincore.org/documents/dcmi-terms/#elements-title
[2] http://dublincore.org/documents/dcmi-terms/#elements-subject
[3] http://dublincore.org/documents/dcmi-terms/#H1
[4] http://dublincore.org/documents/dcmi-terms/#terms-subject
[5] http://lists.w3.org/Archives/Public/public-swd-wg/2009Jan/0000.html

-- 
Tom Baker <tbaker@tbaker.de>

Received on Saturday, 20 June 2009 09:19:28 UTC