Re: Comments on Section 3

"David R. Karger" wrote:
> 
> I don't think that the placement in context changes the fact that
> there is but one "a R b" statement.  I do draw a clear distinction
> between a statement and an _assertion_, which I think is close to your
> notion of stating but which, rather than being about context, is about
> some entity actually declaring that statement to be true.  

I think we are talking about the same thing. I probably shouldn't have used the
term "context", its too overloaded, that was lazy of me. 

By the way "stating" is not my notion, it dates back at least to here:

  http://ilrt.org/discovery/2000/11/statements/

I was just trying to offer it as a possible terminology that might help relate
this discussion to previous discussions on this topic. Given how
confused/confusing/long such discussions have been this might have been an error
on my part!

>    The WG decided that an RDF reification represents a stating rather
>    than a statement, so there can be more than one for each "a R b"
>    triple.
> 
> This seems problematic.  I had been told that one of the points of
> reification was to let someone talk about a statement without stating
> it; are you now saying the opposite?

Not at all. I can have in my knowledge base the statement "Joe said (a R b)" or
more precisely:

   s1  rdf:type rdf:Statement .
   s1  rdf:subject a .
   s1  rdf:predicate R .
   s1  rdf:object  b .
   s1  eg:saidBy  Joe .	        // OK, this one is oversimplified

without having the statement (a R b) asserted in it anywhere. I might believe
that somewhere in the world there is a knowledge base or file into which Joe has
placed the statement (a R b), but I might be wrong, and in any case that doesn't
make (a R b) true in my knowledge base.

All the WG decision means is that if you have a second set of assertions:
   s2  rdf:type rdf:Statement .
   s2  rdf:subject a .
   s2  rdf:predicate R .
   s2  rdf:object  b .

Then you should NOT deduce (without other evidence):

   s2  eg:saidBy  Joe .	

So that s1 and s2 are different statings of the same statement (a R b) and the
property eg:saidBy is talking about the stating not the platonic ideal of the
abstract statement.

> Yes, and I think the URI of the reified statement should be an MD5 of
> the statement.  This means I think that there is only one name to
> represent the reification of a given statement.  But I don't see this
> as a problem.  It is perfectly reasonable for two different
> repositories to contain assertions about a given statement, assertions
> that may contradict one another.

There are lots of times where having a single preferred reification for a
statement is useful and an MD5 hash [*] is a good way to generate a URI for it.
Using the same URI in different repositories is indeed fine. I'm perfectly happy
with this design pattern. You just need to be aware of the consequences of
merging the repositories.

However, because of the WG clarification on reifications it is *also* possible
for someone else to use different URIs for two different reifications of the
*same* statement and to know that conformant RDF processes won't arbitrarily
merge their statements about those two reifications. There are applications
where this is seen as necessary.

Dave

[*] bNodes do cause some problems with this!

Received on Tuesday, 15 April 2003 12:34:13 UTC