Scenario:
metadata-set A describes a digitized version of a book.
In metadata-set A, DC.Source refers to metadata-set B.
metadata-set B described the book.
The problem of indicating that you are linking to a description
of an actual object instead of linking directly to the object
(in other words where does the URI point to?), it seems, requires to
qualify source as surrogate.
In DC-6 there was a consensus that DC.source is a specific type of
DC.Relation. The problem above then would require to specify DC.Relation
first as type=source and then to specify that this "source" is of
type=surrogate and scheme=DC.
How does that translate in syntax terms?
I think the 1:1 model in RDF/XML can accomodate this, but that we as
humans can't "degrade" ourselves to translate this systematically into 1:1
relationships. We will need tools that do the job for us.
We will fill in fields like: author, publisher, source etc...
and the DC-application will translate this into agent.author,
relation.source, etc...
I am convinced that applying and encoding DC according to the 1:1 data
model in RDF/XML will not be humanly possible. Intelligent interfaces will
need to translate complex human metadata concepts into the simple machine
understandable code.
Programmers do not write binary code either. They have very sophisticated
tools.
Cataloguers will need a higher-metadata language that can translate to
DC/1:1/RDF/XML.
gr., Titia
On Sat, 7 Nov 1998, Alex Satrapa wrote:
>
> In order to overcome the problem of linking to a parent/relation's metadata set
> instead of the actual object, perhaps we could do something like:
>
> <META NAME="DC.Source" SCHEME="URL" CONTENT="(type=surrogate)
> http://www.my.domain/surrogate/eiffel_tower_dc.html">
>
> And define the "surrogate" type as indicating that the target of the DC.Source
> link is a surrogate metadata set for something that cannot be accessed
> electronically.
>
> Alternately, we would need to find a way of "linking" to physical objects in such
> a way that we know where to find the metadata for it (if such a set exists).
>
> I have yet to be swayed from the "1:1" argument - which from my perspective is
> simply about what gets assigned it's own standalone metadata set and what
> doesn't. Some physical objects will *need* their own surrogate metadata holder,
> because they'll be referenced from all over the place. I'm sure there's more than
> one work out there which references the Eiffel Tower in some way, shape or form.
>
> Which reminds me - the Eiffel tower was originally erected as a center piece for
> some exposition, wasn't it? If my understanding of history is correct, the Eiffel
> tower was supposed to be a temporary structure? Much the same way that Dublin
> Core was supposed to be an experimental/under development cataloguing tool...
>
> Regards,
> Alex Satrapa
>
> Titia van der Werf wrote:
>
> > On Thu, 5 Nov 1998, David Bearman wrote:
> > >
> > > Clarification of the implementation is something all of us agree is needed.
> > >
> > > I believe that the concept of "record" is and has been getting in the way
> > > of clarity here. In Helsinki we banned uch talk, using set instead. The
> > > operation significance is that some group of metadata elements need to
> > > describe the thing in hand. Another group can describe another thing. They
> > > are linked by relation. The examples I've seen of people whop are doing
> > > what they think is not 1:1 are ok as 1:1 except for not pointing to the
> > > other set. This is to say they contain two groupings of metadata (generally
> > > in one 'record') but don't point to each other.
> > >
> > Hi,
> >
> > Just back in Holland, to support this observation. The problem has
> > also been alluded to as inline or external metadata. This too is not
> > relevant for the 1:1 discussion. In either case you can point to it (well
> > in rdf/xml anyway, as I understood it). What makes 1:1 workable (for
> > machines) is that there is no ambiguity.
> > The example I gave at DC6 with source (nesting of metadata sets!) is not a
> > violation of 1:1, because it can easily be changed into pointers to those
> > metadata-sets.
> >
> > Inheritance of metadata (child/parent) through the
> > relation(IsPartOf) is also an interesting case. I beleive it is not a
> > violation of 1:1.
> >
> > What is a little disturbing however is that (for physical objects at
> > least, but also for electronic resources) we (I mean implementors)
> > actually want to point to the metadata and not to the resource. Strictly
> > speaking then, the URI does not point to the resource that is object of
> > the stated relationship.
> > In DC speak and documentation it is explicitly understood that pointers
> > (URIs) point to the resources and not to metadata about the resources.
> > This is probably cosmetics - but I still wanted to raise this issue here.
> > It disturbs me.
> >
> > regards,
> > Titia
>
>
|