On Fri, 7 May 2004, Lorna M. Campbell wrote:
> Hi Chris,
>
> >And if we are talking about writing metadata for a web page ('with many
> >types')... I would think you should be writing a separate LOM for each
> >image (or whatever) on the page, each with it's own Format entry, and
> >using LOM.Relation's to associate them to a separate LOM for the HTML
> >page itself (hence the existance of the 'ispartof' Relation.Kind). (I
> >asked Wayne Hodgins about exactly this yesterday when he was here, and
> >he confirms this view).
>
> I'm afraid this does strike me rather as an "ideal" rather than a "real" world
> solution. I don't believe that creating an individual LOM record for every item
> that makes up a web page is a real, or indeed desirable, option for most
> metadata creators. And I'd be more than happy to discuss this issue with Wayne
> next time I see him! :-}
I completely agree. The fundamental issue here is agreement on what
*resource* is being described... in the case of most Web pages, I would
suggest, the resource that the end user wants is the 'Web page' - which is
itself a little bundle of objects (HTML file, several GIF images, etc.) -
some of which may be subject to content negotiation, some not. What is
being downloaded to the user's browser is that bundle (or more formally in
Web architecture terms, a set of representations of the resources that
make up the bundle) - that is what the end-user wants to discover.
Therefore, that is the right level of aggregation for stuff to be
described at.
Now there will be some exceptions to this - a Web page containing a very
significant graph for example. The graph may be useful on its own. In
which case the graph can be described sepatately and its relationship to
the Web page (the bundle) documented. But you don't need or want to be
describing every part of every bundle in every case.
Now, at the machine level, it is highly desirable (but not mandatory! ;-)
) that the machine knows in advance what formats are contained in the
bundle - in order that it can predetermine if it is going to be able to
display the whole bundle to the end-user - but at the human level, we are
primarily interested in the bundle.
And, as has been stated before, you may not know in advance what set of
representations of the resources in the bundle are going to be served to
any particular browser.
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/
ECDL 2004, Bath, UK - 12-17 Sept 2004 - http://www.ecdl2004.org/
|