Thank you very much for the excellent replies which I have found very
helpful. In certain cases I have found the responses rather puzzling.
>1. why is the link to the information resource outside the metadata set?
The consensus response here is that DC.Identifier serves that
purpose. But surely the identifier IDENTIFIES the resource it does
not locate (aka link-to) it. Thus a sound metadata scheme
distinguishes between the identifier and the locator or perhaps the
link to the identifier and the link to the service which resolves the
physical address.
By the way, some replies confused this question with the issue of the
location of the metadata. Another matter altogether, covered by the
Warwick Framework.
>2. if all elements in the metadata set are optional then is an empty
>metadata set valid?
The responses were contradictory. Some said yes. Some said no. Quite
a few said who cares. Well ... I care. Call me a picky programming
type but I like to get my specifications correct.
>3. if format is defined by MIME Media Types then why do the list of
>DC Resource Types substantially overlap with it?
The argument here, which I find convincing is that the the overlap is
necessary. I guess a tidy mind would prefer some symmetry between the
top-level media types in MIME and the resource types. Some overlap,
some don't. The same names have different definitions.
>4. how does the format relation type (isFormatOf and HasFormat)
>relate to the DC element source (the definitions appear identical)?
This question opened a can of worms. I found the Ansel Adams example
interesting but not assisting in providing a definition which could
be operationalised (or formalised). I, as do other respondents, think
this issue points to a structural flaw in DC. At any rate the obvious
overlap in this case should be dealt with
>
>5. if elements can occur any number of times why do none of the
>examples in the Guidance on Expressing DC in RDF show this happening?
Perhaps this is a minor quibble. I certainly did not mean to point to
that document alone - its just the most precise and authoritative.
Its just that no examples in any of the literature that I have seen
have repeating elements and I would like to see how they are handled.
I just get a bit worried when everybody says "of course ..." but
there are no examples.
>6. the OMG have developed the UML (& XMI) which supports the
>definition of relationships between objects, how do DC relations fit
>into that picture?
>
I was not proposing defining DC in UML (actually a very good idea,
perhaps I will do it). Rather I was suggesting (in a sort of hesitant
way) using the UML metamodel to define relations (with all that this
brings with it). I think this is an important area worthy of
discussion.
>7. if I make a resource available am I the publisher? Thus if I
>place a resource on the UCL web site it gains its authority from the
>fact that it is located at that domain and hence UCL is the
>publisher lending its imprimatur to my work, even though I made it
>available, or have I misunderstood?
>
There has been much erudite discussion of this which reduces my worry
about raising the issue. It seems to me that "making a resource
available" is not the same thing as publishing and that we can expect
the distinctions in this area to get greyer. My concern is using DC
in an intranet setting where the concept of publisher is very
difficult to interpret. Any advice?
___________________________________________________________________________
Anthony Finkelstein (Prof.) | TEL: 44 171 380 7293 (Direct Dial)
University College London | FAX: 44 171 387 1397
Dept. of Computer Science | EMAIL: [log in to unmask]
Gower Street, | http://www.cs.ucl.ac.uk/staff/A.Finkelstein
London WC1E 6BT | OFFICE: G20, Pearson Building
United Kingdom | PGP Key on request
___________________________________________________________________________
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|