Re: Chainsaw?

At 09:15 AM 11/2/00 -0500, Jonathan Borden wrote:
>     For the type of work we wish to do it is essential that properties be
>assigned to classes.

Do you need properties to be assigned *directly* to classes?  If so, maybe 
we are studying different (but related) problems?

[...]
>     You point out the distinction between a statement made about a class and
>a statement made about the reified statements of a class. Perhaps these
>statements can be modelled as a Bag and assigned an ID.

In a sense, that is almost what happens with contexts as I see them.  (See 
below)

>     For example do you mean to assert: "FordEscord" is definedBy
>"FordMotorCompany" or
>"The definition of 'FordEscord' is definedBy "FordMotorCompany" e,g,

I mean the latter.

><Bag ID="XXX">
><Statement>
>     <predicate resource="rdf:type"/>
>     <subject resource="#FordEscord" />
>     <object resource="rdfs:Class"/>
></Statement>
>     ...
></Bag>
>
>do you mean:
>
><Statement>
>     <predicate resource="definedBy"/>
>     <subject resource="#FordEscord" />
>     <object resource="#FordMotorCompany"/>
></Statement>
>
>or
>
><Statement>
>     <predicate resource="definedBy"/>
>     <subject resource="#XXX" />
>     <object resource="#FordMotorCompany"/>
></Statement>
>
>Assuming we agree that the second statement captures the intent, is this an
>acceptable solution?

My interpretation of the second statement is that it says the *bag* 
containing the statements that define a [FordEscort] is defined by 
[FordMotorCompany].  My goal is to make that assertion about the (reified) 
statements themselves.  (Consider, there may be another bag defined by 
another party containing some of the same statements.)  The nearest thing 
in the RDF spec is <Description 'aboutEach=...'>, but I find that lacks a 
corresponding representation in the RDF abstract model.

As a secondary matter I have a problem with the way that rdf:Bag is 
represented in the abstract RDF model when used in an "open world" 
environment.  When adding a new statement to a collection, it is, in 
general, not possible to know what property identifier should be used 
because one doesn't necessarily know which of the _1, _2, etc property 
names have already been used.  (I think this is noted on DanB's issues 
page.)  Therefore, contexts as I envisage them are at the simplest level 
just a collection of statements, a bit like a bag.  Maybe what we're 
talking about isn't as far apart as may seem.

#g

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

Received on Thursday, 2 November 2000 11:19:37 UTC