Re: Concept spaces and Namespaces RE: Glossary of terms ...

Salut Bernard and all,

I think that it is not a good idea for the SKOS specification to try and
equate a namespace with a set of concepts described as a "scheme". If I want
to create a particular list for a particular purpose, it seems to me that one
of the glories of the RDF-based semantic web is that I can do this simply by
mixing and matching stuff I find, adding my own things only if necessary.

A couple of times I have gone through this exercise in RDF. It is not yet
easy, although tools like Ontaria (which has a smart approach to helping
users) and Swoogle (which has a huge amount of useful data) are starting to
make it a reasonable proposition. I think that we should be encouraging work
in that direction, rather than the continual creation of new URIs that only
serve the purpose of syntactic simplifaction, thus encouraging the
hand-coders at the expense of the tool users (who are the vast majority of
potential users as far as I can tell).

cheers

Chaals

On Tue, 18 Jan 2005, Bernard Vatant wrote:

>2. Not so good idea, since it is at risks to conflate with the "namespace" notion. BTW
>current SKOS specification seems completely agnostic about the relationship between
>namespace and concept space. OTOH, most Thesauri and vocabularies are defined as "unique
>name" spaces. That means legacy concepts will have generally been identified by a unique
>name *inside the concept space* before being ported to the RDF format. So an obvious
>migration practice will certainly be to use a single RDF namespace to somehow represent
>the concept space. I don't know if that should be recommended by the specification, or
>pointed out as a current/best/recommended practice. In any case, I don't think SKOS
>specification should be silent on that point, the more so if it actually shifts from
>"concept scheme" to "concept space".
>
>Just an idea : if indeed it's a good practice to attach a namespace to a concept space,
>why not add this attachment as a deicated SKOS property? It would make useless the
>repetitive declaration of "inScheme" properties, OTOH it does not seem to be consistent
>with the current cardinality of "inScheme" property ...