On Sun, 1 Jul 2001, Chris Croome wrote:
> Hi
>
> On Sat 30-Jun-2001 at 04:35:10 -0400, Dan Brickley wrote:
> >
> > On the basis of these experiments, I'm inclined to advise that the
> > spec be amended to discourage over-use of the RDF container
> > constructs. I'm actually tempted to claim that we should consider
> > avoiding them altogether, since the value of data-merging imho
> > outweighs the information supplied by employing the Bag/Seq/Alt
> > machinery. But I want to make some more test cases before making
> > such a claim...
>
> Point taken, I've removed these.
It was as much a comment regarding a clarification to the spec; but I
think the simplification makes sense.
> I can only think of one thing that I actually _need_ to use a
> container for -- when I want to list something in order.
Yes; RDF Bags (as values of the main DC properties) don't seem a
particularly useful construct. And rdf:Alt is just plain confusing.
Sequences are often useful, though I'm coming around to a view that
they're best useful in custom constructs designed for talking about
ordered lists, rather than being mixed in where you'd expect 'normal'
data. In other words, I don't like having to check each time to see if
dc:creator points to a string, a thing, or a list of things. I can live
with the first two, but to have containers in there as well is making my
DC/RDF code rather complex.
I found Seqs essential when hacking on RDF sitemaps; the oldest RDF
sitemap format was in the earliest release of Mozilla ('Netscape 5' at the
time). This didn't use rdf:Seq and consequently a sitemap viewer app
working with the RDF didn't have access to ordering information from the
original file.
I've just spend a little while tidying up another testbed RDF dataset, and
dug out some examples (screenshots and .rdf files) from the original
Mozilla RDF sitemap format, alongside some test data from an ILRT website:
http://ilrt.org/discovery/2000/08/bized-meta/
This might be of interest, both historically, and as a dataset that can
help us understand when it makes sense to use RDF:Seq for website
metadata...
>
> The child documents of any MKDoc document have 4 ordering options:
>
> * By Title, alphabetical order
> * By date of creation, newest first
> * By date of modification, newest first
> * Manually
>
> How could I represent the order that is selected other that using
> Seq and a list of the child documents?
rdf:Seq seems fine for these, though I think we should find a way of
allowing the basic relations that hold between pairs of documents (next,
previous etc) to not be obscured by additional information supplied using
rdf:Seq.
Dan
>
> Chris
>
> --
> Chris Croome
> http://www.webarchitects.co.uk/
>
|