RE: Provenance in RDF

We use the Jena API also - the URIs are constructed using the Stanford
digest algorthm, or at least a variant extended to include the xml:lang
information.

IMHO if statements have to be reified to be the subject of provenance
attribute, then this would present an intolerable level of complexity.


regards

Nigel Hutchison

Nigel W.O Hutchison
Chief Scientist 
Software AG
Uhlandstr 12,D-64297 Darmstadt, Germany
+49 6151 92 1207




-----Original Message-----
From: Dave Reynolds [mailto:der@hplb.hpl.hp.com]
Sent: Wednesday, February 27, 2002 7:11 PM
To: Hutchison, Nigel
Cc: RDF Interest (E-mail)
Subject: Re: Provenance in RDF


Agreed.

Jena can also do this as it happens - it currently treats Statements as
subclasses (in the java sense) of Resources and so can have URIs and can be
explicitly referred to. This is controversial.

I call this "out of band" because, whilst we can do it at the programming
level
easily enough, it is hard to reflect it in the existing RDF/XML or Ntriple
syntax other than by reification. The nearest offcial syntax support we have
is
"bagID" which not only implies reification but adds an extra bag
indirection. 

Dave

"Hutchison, Nigel" wrote:
> 
> Another way of doing out of band provenance would be to treat the
statement
> itself as a resource.
> 
> Suppose every statement had a URI (U say)
> 
> then we have
> 
> U===> subj --pred--> obj   (U references this statement as a resource)
> for any provenanced (is that a word? :-) values use:
>    U --pv:creator--> "Dave"
>    U --pv:date--> "27/2/02"
> 
> The RDF API would have to have a method that returned the (unique) URI of
> each statement.
> 
> Or is this totally out of band .-) It should work ok with our RDF
> implementation but that's no excuse
> 
> regards
> 
> Nigel Hutchison
> 
> Nigel W.O Hutchison
> Chief Scientist
> Software AG
> Uhlandstr 12,D-64297 Darmstadt, Germany
> +49 6151 92 1207

Received on Wednesday, 27 February 2002 13:34:40 UTC