Re: pls keep Concepts and documents disjoint

* Dave Reynolds <der@hplb.hpl.hp.com> [2004-09-21 14:51+0100]
> 
> Actually use of fragment identifiers doesn't necessarily require you to put 
> the whole thesaurus at a single URL. For example, you could use an 
> arbitrary fragment ID to reinforce that you are referring to a concept 
> rather than a document describing the concept but still put each concept 
> definition at a separate base URL:
> 
>    http://my.org/knowlegebase/chemistry/water#concept
>    http://my.org/knowlegebase/chemistry/ice#concept
> or
>    http://my.org/knowlegebase/chemistry/water#Water
>    http://my.org/knowlegebase/chemistry/ice#Ice

That's true, and DanC has made similar arguments promoting that model 
when we were discussing the Wordnet namespace I made, 
http://xmlns.com/wordnet/1.6/Water etc...

Problem is that it is really pretty awful reading and writing RDF/XML in
that idiom. Perhaps an alternate syntax could make it more bearable, but
you currently need a distinct namespace declaration for each term used.
This hurts my class-based wordnet hyponym vocab more than it hurts SKOS
or the current SWBP wordnet design, since we get to make quite pretty
looking XML with category words appearing as element names. I know
RDF's designed for machines, but the 'RDF syntax sucks' objection is
heard so frequently that I find it hard to ignore, from an outreach
point of view.

> 
> However, I would suggest that the SKOS guide should be neutral on all this 
> allowing developers to chose the form of URI that best fits their needs. It 

RDF vocabs are inherently neutral regarding URI form, or even URIs vs
bnodes, in instance data. That said, is good to have clear examples up
front in docs, and people always copy from examples, so this might not
be entirely sidesteppable.

> would make sense to have a section or appendix referring to the controversy 
> over use of slash v. hash so that developers are aware of the issues and 
> options. However, it doesn't seem appropriate for SKOS to mandate one 
> approach above the other given there are reasonable arguments and existing 
> practice on both sides.

+1

Dan

Received on Tuesday, 21 September 2004 14:06:42 UTC