Contexts and statements, reification and model

At 08:35 AM 11/29/00 -0800, Seth Russell wrote:
>Well, If I understand you correctly,  I think you have put your finger on
>something;  and it is precisely the point where I disagree with Graham's
>definition of what context is.   He says  "context is a collection of 
>~reified~
>statements"; I say "context is a collection of statements".

Seth, you may be right here... I suspect I may have been confusing the 
syntax with the (logicians' sense of) model.

The graph-syntax I propose for contexts has a context resource and graph 
arcs to a number of statement resources.  This syntax is used to represent 
a thing called 'context' that 'quotes' and/or 'asserts' a number of things 
called 'statements'.  If I am understanding correctly, these things called 
'contexts' and 'statements', and the relations 'quotes' and 'asserts', are 
elements of the "model".

Loosely, then, one might say that the context contains the 
statements.  Distinct from that model, the graph syntax uses a resource of 
type context and a number of resources of type statement, the latter being 
reifications of the statements.

>     So the statement
>[id1, s1, p1, o1] can be in the collection, and any statement about that 
>statement
>[id2, s2, p2, id1] may also, independently, be in that same collection (or 
>not) ..
>or any other collection for that matter in which it belongs.  So when 
>we  browse
>our attention to a particular context, the statements in that context show up
>...that is [contains (idX, sS, pX, oX)] becomes true.  I mean, what you 
>say about
>what I say, certainly is not necessarily in the same context as what I say, or
>even what I mean.

Here, I think it may be you who are confusing syntax and model;  if I 
understand correctly, the 'id1' and 'id2' are artefacts of the graph 
syntax, not part of the model that includes statements, etc.  Thus, I would 
rephrase your first sentence above to read:

   So the statement [s1, p1, o1] can be in the collection, and any statement
   about that statement [s2, p2, [s1, p1, o1]] may also, independently,
   be in that same collection (or not) .. or any other collection for that
   matter in which it belongs.


[Just re-read Sergey's posting of Tue, 28 Nov 2000 13:58:52 -0800...]

Once I separate the model from the graph syntax in this way, I think it is 
also consistent with what Sergey is saying about treating statements as 
resources within his implementation.

Armed with this new understanding, I've just reviewed the formal model 
section in RDF M&S, and most of it seems very straightforward, except this:

>9. Reification of a triple {pred, sub, obj} of Statements is an element r 
>of Resources representing the reified triple and the elements s1, s2, s3, 
>and s4 of Statements such that
>
>       s1: {RDF:predicate, r, pred}
>       s2: {RDF:subject, r, subj}
>       s3: {RDF:object, r, obj}
>       s4: {RDF:type, r, [RDF:Statement]}

I note:  this clearly states that the reification of a statement is a 
resource.  If one accepts that the identifiers do not appear in the formal 
model, then I think the reified statement must be unique, per Sergey's 
comments.

This seems to be at odds with a previous statement I made 
(http://lists.w3.org/Archives/Public/www-rdf-interest/2000Nov/0334.html) 
in  which I say:

>As I follow this debate, I become more convinced that the idea of a
>reification as a "model" of a statement is most helpful.  It allows us to
>create such a model, with associated resource ID, so we can make statements
>like "A statement of the form [p s o] was asserted and signed by Alice",
>and also "A statement of the form [p s o] was asserted and signed by
>Bob".  Here, we actually _need_ two separate resources to express this
>without also saying "A statement of the form [p s o] was asserted by Alice
>and signed by Bob".
>
>Thus, I suggest, we actually need to be able to have multiple resources
>representing a given statement.  The RDF model achieves this through
>reification.  The XML syntax and any implementation may provide a shorthand
>for the reification quad.

Reflecting on this, I think the weakness in that earlier statement was in 
"A statement of the form..." as opposed to just "the statement".  If I need 
to talk about a particular stating being asserted and signed by Alice, and 
another stating asserted and signed by Bob, I think I need to create two 
new resources that relates to the (unique) reification of [p s o].

This seems to go against Seth's original comment that prompted this 
message.  Currently, the RDF model requires models that talk about 
statements to contain _reified_ statements.  I think either approach could 
work, but the spec currently uses the reified form.

#g


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

Received on Thursday, 30 November 2000 12:33:02 UTC