Re: A simple question....

Stefan Decker wrote:

> To make a more general point:
> If one views RDF as a simple Frame system, then there are two things which
> distinguish it from other frame formalism:
>
> 1) It is distributed in an uncontrolled fashion
> 2) It is linked to Web-Resources
>
> both aspects have to be handled very carefully, because they introduce
> something new to well known representation formalism.
> E.g. 1) Introduces the aspect, that metadata has a source, and there might
> be sources i consider  more reliable than others.
>             So i want to be able to distinguish between metadata from
> different sources.
>             RDF datamodel does not allow that (the datamodel does not trace
> the origin of a triple)

I deal with this by having my Arcs (triple) remember thier "home" ArcSet (my
concept for the arc containers that are called Models in most of the RDF
discussions).  Arcs can be incorporated into more than one ArcSet but they
remember their home. Furthermore my ArcSets remember all the ArcSets that are
directly aggregated into them. I am hoping that this allows enough traceability to
meet my needs.

> The second aspect (links) means, that it has to be declared, what it means
> to link to something on the web.
> and to distinguish between different types of links (which are useful in
> applications).

My tendency is to place the interpretation of the links on the application. I'm
less worried by how to interpret the links than how to figure out if different
nodes are equivalent. I.e. the "==" vs. "equals" issue but for URI. This seems
like a much bigger problem than the  equivalence of the URI and the value of the
dereferenced URI.  OTOH, I may be completely missing the point :-)

Cordially from Corvallis,

Gabe Beged-Dov
http://www.jfinity.com/gabe

Received on Wednesday, 17 November 1999 19:27:06 UTC