RE: lightweight reification (was Representing trust (and other context) in RDF)

Brian,

I can't speak for Dan and Guha, but this is exactly the kind of thing that 
I am chasing when I talk about associating context with 
statements.  Especially your final diagram below.

And I would be more than pleased to find a way other than digests or full 
reification to do this, but I cannot see such a way that works within the 
current RDF model.

#g
--

PS:  I'd also observe that the trickiness of this concept "implied in the 
model and syntax document" is much clearer when one thinks primarily in 
terms of the graph model rather than its XML serialization.  The M&S 
document addresses it through reification, but...

--

At 10:27 AM 5/29/00 +0100, McBride, Brian wrote:
>It had occurred to me a little while back that there was a concept implied
>in the model and syntax document, but not addressed explicitly in it.
>
>When we talk about RDF models or RDF serializations, there is an implied
>container that the specs don't talk about.  So when we write a model like:
>
>     [subject1] ---[predicate 1]---> [object1]
>     [subject2] ---[predicate 2]---> [object2]
>
>There is an implied 'model' or StatementSet of which these statements are a
>member.  This is a concept that is clearly present in Sergey's API.  Each of
>these statements has the implied 'property' that it is a member of a model.
>
>It is useful to be able to assign a URI to a model so that it can be
>identified and so that statements may be made about the collection.  This
>might be represented as:
>
>    +--Model------------------------------------+
>    |                                           |
>    |   [subject1]---[predicate 1]--->[object1] |
>    |   [subject2]---[predicate 2]--->[object2] |
>    |   [Model]---[asserted by]--->[Brian]      |
>    +-------------------------------------------+
>
>Again this is what Sergey has proposed, though I think there are
>alternatives to digest URI's to identify the model.
>
>This is certainly useful when building a large repository.
>
>When statements are treated as resources, there can be models of the form:
>
>         +--[subject]--[predicate]-->[object]
>         |
>         |
>    +----|--Model-------------------------------+
>    |   \ /                                     |
>    |   [subject1]---[predicate 1]--->[object1] |
>    |   [subject2]---[predicate 2]--->[object2] |
>    |   [Model]---[asserted by]--->[Brian]      |
>    +-------------------------------------------+
>
>which is intended to indicate that subject1 is the statement outside of the
>Model.  Thus a model can contain statements about a statement S, without
>containing S itself.
>
>At the time it also seemed this might provide a way of ducking metaphysical
>issues of 'belief' and 'truth' and reducing them to statements about set
>membership, but I really must go chase those pointers from Dan and Guha as
>there has clearly been much deeper thinking about this than I have done.
>
>Brian

------------
Graham Klyne
(GK@ACM.ORG)

Received on Monday, 29 May 2000 18:04:46 UTC