RE: Representing trust (and other context) in RDF

Brian,

What I am trying to accomplish is to find a *representation* (as opposed to 
implementation), within the RDF model (as opposed to serialization), for 
trust-related information.  Who signed what statements, etc.  I think this 
is a general problem, so it is appropriate to seek a "standardized" form 
for representing such information rather than to have each application of 
RDF invent its own solution.

I think we may be trying to say the same thing, or something very 
similar.  Certainly, Guha's comments about representing contexts as 
resources seem entirely consistent with what I was trying to suggest.

Unfortunately, I find XML serialization of RDF is very difficult to read, 
and not helped by the HTML formatting of your message, so I can't tell 
exactly what you propose in your message.  It looks as if you are 
suggesting something similar to Sergey's proposal that I reference from my 
original posting.

Thus, if there is a difference between our proposals, I think it is that I 
am seeing a 'context' as a property of a statement, where you seem to 
suggest a statement is a property of a 'statementset'.  In either case, we 
need a way to represent a statement (preferably without invoking the 
machinery of reification).  So, an additional ingredient in my suggestion 
(and Sergey's) is the use of a digest function to represent a 
statement.  (The details of how to compute a digest are TBD -- and, yes, 
anonymous resources may pose a challenge.)

I would find it easier to understand your thinking if you could show your 
view of an RDF graph representing a collection of statements that have been 
signed by an identified person.

#g
--


At 06:53 PM 5/24/00 +0100, McBride, Brian wrote:
>I'm not sure what you are trying to accomplish here Graham.  If you are
>looking for an implementation, consistent with the current RDF specs that
>will do the job, then I don't think any extra machinery is needed.  Have a
>look at the bit about StatementSet's in:
>
>http://lists.w3.org/Archives/Public/www-rdf-interest/2000May/0013.html
>
>
>This is I think just what Guha has described.
>
>I'm also not sure how one computes a digest for a statement which refers to
>an anonymous resource
>
>If you are looking to extend the model to support context's without the
>percieved bloat that results from reification, then could one just declare
>statements to be resources.  This is in fact what reification does, but
>might be a more palatable way of explaining it.
>
>Brian
>
>
> >-----Original Message-----
> >From: Guha [mailto:guha@guha.com]
> >Sent: Wednesday, May 24, 2000 17:26
> >To: Graham Klyne
> >Cc: RDF interest group
> >Subject: Re: Representing trust (and other context) in RDF
> >
> >
> >The way the AI folks have been modelling contexts ---
> >
> >if source--arc-->target (written as arc(source, target))
> >is in context C, you write it as ist(C, arc(source, target)).
> >"ist" is read as "is True In".
> >
> >C is itself a first class object (i.e., a resource). The collection
> >of statements that are true in C could be closed under
> >deduction.
> >
> >The common frameworks for contexts allow for lifting
> >of statements from one context to another. i.e., if a statement
> >P is true in C1, one can conclude that it is true in C2.
> >
> >Guha


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

Received on Thursday, 25 May 2000 06:49:05 UTC