[MOBY-dev] RDF descriptions of MOBY services

[Mark Wilkinson]

Hi all, 

Nina and I are debating the most appropriate way to represent a MOBY
service signature in RDF.  The main source of controversy is with the
inputs and outputs, in particular when it comes to collections.

In the current draft, I have the input as a bnode type:Bag, with
individual input articles coming off of it.  Input articles in MOBY are
either Simple or Collection, however at the moment I have made this
distinction implicit - if the article is type:Bag, then it must be a
Collection, and if it is not type:Bag, then it represents a Simple, and
has various information (object_type, namespace, etc.) predicated to it.

What concerns me is this:  If we make the interpretation implicit, then
we don't need additional predicates, whereas if we explicitly define the
edge to an article as "has_simple", or "has_collection", then these
predicates need to be defined.  The data structure (Bag) is semantically
identical (IMO) to what a Collection article is in MOBY, so I don't see
a *need* to explicitly define this... but under either circumstance the
person designing a query or a parser would need to understand how to
interpret either the "shape" of the graph (if we leave it implicit), or
the meanings of the predicates (if we make it explicit).

...??  This never ceases to befuddle me.  It seems that, even with RDF,
we can not get away from the requirement of community agreement on
interpretation.  I guess this reminds me of Phil's recurring statement
"ontologies are only useful if they are shared"...

...but from those of you with more experience in RDF than I have, could
you advise me which of these two options is "more correct"?  Is it
better to be explicit, and define new project-specific predicates, or is
it better to let the data structure implicitly speak for itself using
existing RDF/S predicates?

Your wisdom is welcome!

M

-- 
Mark Wilkinson (mwilkinson at mrl.ubc.ca)
University of British Columbia iCAPTURE Centre