On Thu, 2 Mar 2000, Sigfrid Lundberg, Lub NetLab wrote:
> > I looked back at the ballot and do not see any proposals from the Agent
> > WG for constructions analogous to DCBOX and DCPOINT. Which of the
> > qualifiers proposed by the Agent WG do you see as a structured value
> > (as opposed to a value component)?
>
> I don't see any difference between structured values and value
> components. Value components are now obsolete and propesed in reports that
> are now seen as faulty by some of their authors. Therefore someone
> proposed structured value instead. Don't know who. It amounts to the same
> thing.
Structured values, in my understanding, are strings that have multiple
components, either without labels ("Thomas Baker, GMD") or with labels
("<name>Thomas Baker</name><affiliation>GMD</affiliation>"). I
understand value components to be the component parts of such values,
such as "affiliation".
The Agent WG has proposed to recognize certain *components*, such as
Affiliation, as qualifiers of the CCP elements. They have not proposed
structured values -- entire templates of components structured in
certain community- or application-specific ways (such as "Last name,
first name (lifespan), affiliation").
> > DCBOX and DCPOINT do cross that fuzzy line beyond which syntax is mixed
> > with semantics. I must admit that I was willing to relax the criteria
> > in the case of Coverage, because it is not clear to me how this
> > element, which can hold many types of data about Place and Time, can
> > gracefully dumb-down at all.
>
> Well, I'm _not_ willing to relax anything. I want a uniform, logical
> structure.
I'm willing to follow you on that.
> A DCBOX HASA (using your terminology) northwestern corner and it HASA
> southeastern corner. There is no difference logically between something
> that has a corner or something alse that HASA date of birth
>
> > If it sets a bad precedent, however, and opens the door to putting
> > structured values about agents into the Core itself, then I think we
> > are right to leave DCBOX and DCPOINT off of the ballot for DCMES
> > qualifiers.
>
> Come on Tom.
>
> I'm not saying that we should move them in. I'm just saying that the DCMES
> has to acknowledge that some values belong to a structured data type. The
> DCBOX as modelled by Simon is already in a seperate name space.
Affiliation can be used in a structured value, yes. And those structured values
are modelled somewhere outside of the core, yes.
> DCBOX is much more useful than a the points that are useless for resource
> discovery.
We agree on this point completely -- DCMI must have a way to recommend
it. However, we may disagree on the form that recommendation should
take. I do not see how structured values as wholes (templates) fit
into our basic dictionary -- the element set. Is there anything wrong
with presenting such templates in the context of user notes (which
could be given Recommended status)? That would leave the Core itself
with a more uniform, logical structure, which is what you want.
Tom
_______________________________________________________________________________
Dr. Thomas Baker [log in to unmask]
GMD Library
Schloss Birlinghoven +49-2241-14-2352
53754 Sankt Augustin, Germany fax +49-2241-14-2619
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|