[sc34wg3] Individual contribution on the U.S. N.B. position o nthe progress ion of Topic Map standards

Robert Barta sc34wg3@isotopicmaps.org
Sat, 3 Apr 2004 20:15:33 +1000

On Fri, Apr 02, 2004 at 09:19:03AM +0200, Jan Algermissen wrote:
> Dmitry wrote:
> > We can define following additional axiom in "extended" theory:
> > 
> > If X1 has BaseName Y1 in scope S
> > and  X2 has BaseName Y2 in scope S
> > Then X1 has the same subject identifier as X2
> So you need to make up an identifier for this and make it a
> subject indicator of both topics. Can you sketch an example
> of how this would look?


AsTMa! can do these things:

# identification by identically scoped basenames
forall [ $a
         bn @ $scope : $bn ]
   => derived exists [ $a reifies $x ]

# the following is actually built-in in the data model, but that is
# the only rule

forall [ $a reifies $x ]
   => not exists [ $b reifies $x ]

> > Now let's say that it is not enough. We can add new axiom:
> > 
> > If X1 has Internal occurrence O with  value  V in scope S
> > and X2 has internal occurrence O with value V in scope S
> > and  occurrence O is InverseFunctionalOccurrence
> > Then X1 has the same subject identifier as X2
> See above.

See above. :-)

> > As soon as data model is TMDM all possible merging rules are
> > expressible as additional axioms which reduce identity management to
> > only one basic  merging rule based on subject identifiers.
> But then you have just moved the rule definition from properties to
> characteristics - where is the gain?

1) We do not burden the _data model_ with these things. I would assume that
   all attempts to handle this gracefully have failed, because it is simply
   not the right place to do.

2) Enourmous flexibility. The above equivalence relation between topics
   is induced by names in scopes. But I can use _any_ rule which is possible
   in AsTMa! to induce other equivalences

> > With RM data model I am limited with syntactic low level constructs
> > such as SIDPs and basic assertions. It is difficult
> > to express rich identity rules at this level (with almost zero
> > semantics) 
> Noone has ever said that the RM is to be used that way. The RM
> enables the definition of the semantics you talk about. THAT is
> its purpose.

Is it declarative?