On Wed, 15 Oct 2003, Douglas Campbell wrote:
> Andy,
>
> >>> [log in to unmask] 15/10/03 10:45:52 >>>
> The problem is that, although the two forms looks very similar, there
> is
> nothing in RDF to say that the first form can be transformed into the
> second form. DCMI can define the transformation quite easily - but
> this
> will not be performed by non-DCMI RDF-aware applications.
>
> I think I agree in principle to 'every value is a resource', but I'm
> not sure
> about how this should be represented in RDF. A lot of RDF data seems
> to be like:
> R1 --predicate--> "literal"
I'm not arguing about whether it is right or wrong to do this. I'm
arguing that it is confusing for DCMI to have two documents (one
recommendation and one proposed recommendation) that recommend different
constructs. Both constructs may be perfectly valid ways of modelling the
world - but there is no widely agreed way of mapping one construct onto
the other. Therefore, I'm arguing that DCMI should use the same
constructs (the same model) in all its RDF-related recommendations.
> I'm trying to understand when it is valid to use a literal at the end of
> a triple and what it states - we are now saying that it should never be
> used like this for DC data and instead literals should always be a label
> of a resource.
That's what I'm suggesting, yes.
> Could it be the case that a lot of RDF data is using literals meaning
> them as labels for conceptual resources (as we used to), which would
> then potentially put this new DC form at odds with the rest of the RDF
> community?
That could be the case, yes. Hence the need for discussion here. But my
suspicion is that you'll find both forms in use within the community - at
least in part because people will have read one or other of our documents
or both and they'll be slight confused about what to do. For example,
there have been various discussions on the RDF interest list about whether
DCMI intends all its values to be 'strings' or not.
> As an aside, the elements I find hard to think of as conceptual
> resources rather than values are dc:title and dc:description.
By which I assume that you mean that these things are always 'strings' of
characters?
Well, what is the dc:title of a story that has only ever been told
verbally? Sounds a bit Zen-like doesn't it! :-)
Or, to put it another way, when I say to you (verbally), "the title of
this book is The Lovely Bones" I'm not giving you a string of characters
- but I am telling you what the title is. So, there is a conceptual
resource, the title of the book, sitting somewhere behind the string of
characters that is the most common way we have of denoting that
resource.
Does that make any sense?? :-)
Andy
--
Distributed Systems, UKOLN, University of Bath, Bath, BA2 7AY, UK
http://www.ukoln.ac.uk/ukoln/staff/a.powell +44 1225 383933
Resource Discovery Network http://www.rdn.ac.uk/
|