[sc34wg3] Question on TNC / Montreal minutes
Lars Marius Garshol
18 Sep 2002 22:28:17 +0200
* Lars Marius Garshol
| You should be aware, however, that many of the applications I have
| written do merging based on names. I still don't think having that
| behaviour in the standard is right.
* Marc de Graauw
| But you never say WHY it should not be in the standard. Your
| position is - if I understand you correctly - top put merging based
| on name, occurrence, association in TMCL, not the SAM.
That is correct.
| But what's wrong with doing this in the SAM? Why is it correct to
| describe merging based on subject identity in the SAM, but not so
| for name-based merging?
That's a question that deserves a good answer.
Firstly, merging by URI should be in the SAM because by the semantics
of those constructs are such that it is clear that there will never be
any exceptions to the rules. That is, if two topics have the same
subject indicator they must have the same subject and no exceptions to
that rule are reasonable. (Another reason is that merging by URI is
simple. The rules is straightforward and require no extra information.)
Secondly, merging by topic characteristics should not be in the SAM
for several reasons:
- one needs a relatively complex machinery for indicating precisely
what the rules are (merge all topics of type X plus subtypes which
have the same base name regardless of scope, merge all topics
which have the same occurrence of type 'email' in the same scope,
and so on). The complexity of these rules make them unsuitable for
inclusion in the core standard.
Most likely we will need to create either a syntax for such rules,
or at least a set of published subjects. Does this belong in the
SAM? I feel it is quite obvious that it does not.
- a demand that the values of a particular occurrence type be unique
is essentially a constraint, and as such logically belongs
together with the other constraints. (This is how RDF does it.)
- having TMCL constraints available makes it possible to improve the
results of merging in several interesting ways. This means that it
makes good sense to have two sets of merging rules: basic merging
by URI in the SAM and improved merging based on constraint
information in the TMCL. (See issue merge-use-of-schemas.)
I hope this answers your question.
Lars Marius Garshol, Ontopian <URL: http://www.ontopia.net >
ISO SC34/WG3, OASIS GeoLang TC <URL: http://www.garshol.priv.no >