Every so often something or someone prompts me to take a look at RDF
(resource relationships), and that look usually expands out to RDFS
(schemas for certain constraints relationships) and OWL (how to
express more generally specific constraints among relationships). The
descriptions in parens are my paraphrases.
In the context of the RESTful-JSON group, Bill de hÓra compared the
discussion there on URLs for JSON object attributes to using URLs in
RDF from a subject to either a literal or another resource (URL).
Which led me to reread Bill's piece from a few years ago on attempting
to automate the linkages between HTML forms and form elements and
updating resources designated in a service using RDF...
http://www.dehora.net/journal/2005/08/automated_mapping_between_rdf_and_forms_part_i.html
This aspect of REST, i.e. the client side being more general, and not
one-off for a specific model, is still lacking from what I have seen,
for RDF as well as now for JSON. For JSON the solution seems to be
send down the Javascript that is aware of the model expressed in JSON.
I think the same would be true for RDF, except when you get into the
need to make inferences, and so on, then more processing is required.
For either RDF or for JSON, modelling conventions are needed to be
more generically expressive in the client per se. Otherwise
conventions are needed server side to map from simple forms clients as
discussed by Bill.
So anyway... now I am interested in RDF and going a bit deeper with it
than before. I see the RESTful JSON development leading very quickly
into discussions of conventions of usage - what is a "collection"?
what is a "reference"? and so on. So I picked up a new good book
reviewed not long ago by Danny Ayers...
Danny Ayers...
http://blogs.talis.com/nodalities/2008/06/first-impressions-of-semantic-web-for-the-working-ontologist.php
Which led me to a comment on that review by Rowland Watkins. He commented...
"I am surprised that RDFS is still being used (except for certain
relationships), let alone RDFS-Plus - mixing RDFS and OWL is not a
good idea (this may be stated in the book). Unfortunately, FOAF still
suffers from RDFS/OWL (making it OWL Full). Still takes a bit of
searching to find the OWL DL version…"
Which catches this non-semantic-web-insider by surprise. A web search
did not turn up anything about mixing these not being a good idea. I
see a lot of discussions about subsets of OWL (lite, DL, etc. as well
as semi-custom) and so on, for decidability, performance, etc. But I
am wondering what the state of the semantic web standards and
implementations really are?