Re: RDF Profiles in Link Values

On 11/17/06, Xiaoshu Wang <wangxiao@musc.edu> wrote:
>
>
> > My €0.02: Ignore the titles, or take them just as a very broad hint.
>
> I would take the same approach too.  Title is supposed to be a hint for
> humans, not machines.  In fact, the combination of the "meta" and
> "application/xml+rdf", suggests an RDF/XML parser.  If you have such a
> parser, it should be able to understand any RDF model.  Then, why not simply
> pointing to one external RDF document.  In this document, use some RDF
> vocabularies to describe the location of different profiles, something like
>
> ex:resource title:foaf <uri_1>;
>               title:sioc <uri_2>;
>             title:doap <uri_3>.


Hints can be useful both for machines and humans. Machines are used by
humans after all.
And if the document itself is not human-readable, what value its title is it
to a human reader?

Adding yet another RDF document does not pass the rule that "everything
should be as simple as possible [but not simpler]". A collection of "meta"
tags do provide links to all the profiles and add some hints. It is hard to
prove necessity for an additional RDF document - it would be like using
rdfs:Bags (which are almost extinct now) when not needed. Using a number of
separate "meta" links with type "application/rdf+xml" as it is now achieves
the same purpose.

I would say that having hints is good (unless there is something as simple
and better).
Yet, we can argue about how to best provide links or hints - which is being
discussed in this thread now.

It would be a mistake to say that this pattern is only now appearing -
autodiscovery links are a well established pattern. It has been around long
enough for some RDF autodiscovery tools e.g. FOAFer [1] to no longer be
working with Firefox 1.5 and for new tools like Semantic Radar [2] and
FOAFox [3] to appear.

[1] http://crschmidt.net/foaf/foafer.html
[2] http://sioc-project.org/firefox/
[3] https://addons.mozilla.org/firefox/2633

Recommendations for its use (or abuse) are integrated in the documentation
of a number ontologies including FOAF [4],  ICRA Labelling [5],  DOAP and
SIOC and are used by various providers of RDF (FOAF, SIOC, DOAP, ICRA, some
RSS 1.0) data on the web, such as LiveJournal FOAF export. Alternative
methods for pointing to related RDF information like using RDFa [6] may take
over in the future, of course.

[4] http://rdfweb.org/topic/Autodiscovery |
http://iandavis.com/blog/2003/02/foafAutodiscovery
[5] http://www.icra.org/systemspecification/#specificLink
[6]
http://b4mad.net/datenbrei/archives/2006/06/07/seealso-for-sioc-hooked-in-page-via-rdfa/

But because it is very simple and so similar to the widely used RSS
autodiscovery (though I do not know if it "abuses" the title or not) it is
probably not going away any time soon. My initial thought was that
specifying ontology identifiers (like Sean proposes) may be of help to
improve this pattern but now I am not so convinced - these links are just
hints and free text works well for that - if an application needs to know
more it shall retrieve these resources and see for itself.

P.S. Yet these hints are helpful for such simple applications like Semantic
Radar where RDF parsing may not always be easy to accomplish or where
browser security settings may prevent access to resources that are not
located on the same host that the user is currently viewing.

Uldis

[ http://captsolo.net/info/ ]