TMCL Uniqueness rule Re: [sc34wg3] Draft Reference Model

[Lars Marius Garshol]

* Lars Marius Garshol
|
| This is what I proposed in Montr=E9al that we should have in general,
| rather than the specific and restricted TNC rule. We certainly do
| need this. The only trick will be working out how to harmonize
| uniqueness rules in TMCL with SIDPs in the RM.

* Sam Hunting
|=20
| OK, I'll bite. What's a "uniqueness rule" in TMCL?

It's just an idea at the moment, but what I was thinking was that
TMCL should allow the expression of rules like the following:

 - merge all topics that have the same value for the 'email'
   occurrence,

 - merge all topics of type 'country' that have the same name in the
   same scope,

 - merge all topics that play the role 'father' or 'mother' in a
   'parent-child' association with the same 'child' in the same scope,

and so on.=20

ISO 13250:2002 says that the identity of subjects may be "inferred
from the topic's characteristics", and that seems to me very reasonable.
It also seems that it should be possible for TMCL schemas to declare,
as part of the constraints on an ontology, the uniqueness constraints,
and for these to be used for merging.

I think probably it is better to call the constraints uniqueness
constraints in TMCL, and then to let users decide whether to apply
them to detect violations or to apply them to do automated merging.

Essentially, this is an idea related to DAML's unambiguousProperty,
but taken one step further.

<URL: http://www.ontopia.net/omnigator/models/topic_complete.jsp?tm=3Dtm-st=
andards.xtm&id=3Dsubject-identity-establish >

--=20
Lars Marius Garshol, Ontopian         <URL: http://www.ontopia.net >
ISO SC34/WG3, OASIS GeoLang TC        <URL: http://www.garshol.priv.no >