Re: Is "Model" part of the RDF model?

Hi Brian,

Based on the discussion and references I guess we can say that a concept of
context is needed. Its certainly vital in our applications that merge
statements from different sources.

At a minimum, a context is a collection of reified statements, although
Graham Klyne's paper takes it much further.  It seems contexts can be
represented within the existing RDF model, even though they are not defined
in the model.  However, it seems the syntax presents problems because you
can't always determine the URI for the reification of a statement.

But what is the proper API support?  To elaborate on the two alternatives:

1 -- An API without the Model concept, or at most a singleton Model
representing the universe of statements.  This API must provide an efficient
means to navigate from a statement to its reification and back.    The
concept of contexts can be built on that, using an appropriate schema but
without any additional specific API support.

(When a statement is first asserted or parsed from a document you would have
to be careful to associate its reification with a context or you might lose
track of it! )

2 -- In the alternative, the Model interface that directly implements a
context.  This might avoid materialising the reification of each individual
statement at a low level. In other words, those resources remain hidden
behind the Model interface. Hopefully the interface could be defined such
that it can be fully understood within the RDF data model together with the
assumed context schema.

The Stanford and Jena API's provide a Model that is almost a context.  Maybe
the connection could be formalised and a core context schema incorporated,
per (2).   Moving in the direction of (1) leaves it all to the application
layer and exposes a lot of fine-grained reification.  We did something like
(1) in our software and I feel the need for more built in context support.

Just my thoughts on it,
Regards
Arnold
--
Arnold deVos
Langdale Consultants
adv@langdale.com.au

----- Original Message -----
From: "McBride, Brian" <bwm@hplb.hpl.hp.com>
To: "'Arnold deVos'" <adv@langdale.com.au>; <www-rdf-interest@w3.org>
Sent: Saturday, October 21, 2000 3:31 AM
Subject: RE: Is "Model" part of the RDF model?


> Arnold,
>
> A good question.  I think the current situation is as you say
> that model is not part of the the RDF Model.
>
> I hope Sergey will respond with his views on the Stanford API.
> Also relevant to this is the idea of contexts proposed by
> Graham Klyne.  His paper can be found at:
>
>   http://public.research.mimesweeper.com/RDF/RDFContexts.html
>
> Jena's precursor was an implementation of a variant of the
> Stanford API with a different approach to handling the URI's
> of models.  It's models did have URI's and that was pretty
> useful. I nearly did the same thing with Jena, but decided
> I wanted to think about it a bit more and put that off
> to a second release.  Jena in its current form is an
> experiment to test out a style of API and is not yet final.
>
> I don't think the API's use of models is a disguised attempt
> to change the model.  I have suggested adopting a variant of
> of the RDF Model which in which models are explicitly defined.
>
>   http://lists.w3.org/Archives/Public/www-rdf-interest/2000Sep/0121.html
>
>
> There is a clear need to be able manipulate collections of
> statements.  Sometimes we want to serialize a particular
> collection into a document, or do a query on a given collection
> or make statements about a collection.
>
> There seem to be two main approaches to doing this.  There can
> be a concept of a collection of statements, i.e. a model aka
> StatementSet.  Or there is only one universal model and RDF
> itself is used to create collections of reified statements
> within it, as bags or contexts.
>
> In Jena's predecessor, models had URI's because I was
> implementing a SQL based RDF store and I needed a way
> to name different collections of statements.  It seemed
> appropriate to use URI's and treating collections of
> statements as resources enabled me to attach properties
> to the collections.  In that regard, I think that those
> models were very like what you are doing with bags of
> statements.
>
> One of the tasks on my todo list is to think about what the
> Jena API would be like if it had no concept of model.
>
> I think where I'm at at the moment, is that if model's exist
> they should be resources.  But I'm still thinking about
> whether they need to exist at all.
>
> Brian
>
>
>
>
>
>
> > -----Original Message-----
> > From: Arnold deVos [mailto:adv@langdale.com.au]
> > Sent: 20 October 2000 08:36
> > To: www-rdf-interest@w3.org
> > Subject: Is "Model" part of the RDF model?
> >
> >
> > RDF API's Stanford [1] and Jena [2] seem to share the concept of a
> > "Model" defined as a "container of triples" or "a set of Statements"
> > respectively and not to be confused with *the* RDF model.
> >
> > There is no explicit concept of a Model mentioned in the formal RDF
> > definition [3], so my question is: is Model just an API mechanism or
> > is it actually an extension to RDF in disguise?
> >
> > I suspect the latter because the  distribution of statements among
> > Model entities may influence the outcome of inferences and queries.
> > For example, one might select among conflicting statements based on
> > their provenance i.e. based on the Model that contains them.
> >
> > Indeed, in the Stanford API the Model has a source URI so it can
> > correspond to an RDF document. But again, is "document" a concept in
> > the formal RDF model or is just part of the syntax?
> >
> > In our in-house RDF engine we implemented "provenance" by reifying
> > statements and putting them in a bag. Statements about the bag
> > influence the credibility of the reified statements for the purpose of
> > inference and queries.
> >
> > In the Stanford API (but not Jena) the Model is-a Resource.  Is this
> > resource intended to be the same as our bag?
> >
> > Any insights into this would be helpful.
> > Regards,
> > Arnold
> > ---
> > Arnold deVos
> > Langdale Consultants
> > adv@langdale.com.au
> >
> > [1] http://www-db.stanford.edu/~melnik/rdf/api.html
> > [2] http://www-uk.hpl.hp.com/people/bwm/rdf/jena/index.htm
> > [3] http://www.w3.org/TR/REC-rdf-syntax/#model
> >
> >
> >
> >

Received on Saturday, 21 October 2000 09:20:05 UTC