[sc34wg3] Association items

Nikita Ogievetsky sc34wg3@isotopicmaps.org
Sat, 11 Jun 2005 14:23:55 -0400


! * Nikita Ogievetsky
! | In any case, I think that real case for unary associations is in the
! | data gathering exercise, rather then in data modeling.
! What's the difference?

When gathering data one should be prepared to encounter things that were not
provisioned in the model. 
When modeling one should provision / predict patterns of data that will be

In fact it is quite possible that there should be two different
sub-standards of topic map: 
- Topic maps for knowledge aggregation (TM4KA)
- Topic maps for knowledge sharing (TM4KS)

! | On the other hand, here is a brief summary of my understanding of
! | the confusion related to this subject:
! |
! | Some people believe that there is no such thing as an "invalid topic
! | map".  Like there is no such think as an "invalid opinion".
! I'm not sure I agree with this. I think the situation for topic maps
! is the same as for XML: If you don't meet the syntax rules it's not a
! topic map (ie: invalid XTM is not a topic map, it's just an XML
! document). Similarly, if the topic map is not valid by schema X, well,
! then it's not valid by schema X, but it's still a topic map, and it
! could still be valid by schema Y.

Reusing acronyms from above,
TM4KA requires all constraints to be maximally removed, 
TM4KS benefits if accompanied by a schema/ontology/etc.

TM4KS schema feeds initial state of TM4KA,
TM4KS once complete leads to a new version of TM4KS schema that then feeds

! | Others believe that topic maps should compete with RDF as an
! | inference-enabling technology.
! I don't think that's a competition we should enter. Topic maps are, by
! their very structure, optimized for a different purpose. You can still
! do logical inferencing on them, but I think one could legitimately
! question whether that's a primary concern for topic maps.
Absolutely agree with this.
! | For some people associations mean "typed" sets, and as sets they can
! | be empty or unary.
! TMDM (and XTM) do not accomodate this view, as associations must have
! at least one role.

Empty association is really a dumb case unless it is some sort of a
singleton (for example if there is a constraint that there can be only one
association of that type within a given context, hmmm... does TMCL provision
for such thing?)

! | For others associations are typed relationships and so there are
! | should be 2+ related "things"/"subjects"/"players".
! I think this is where Patrick started from. I guess he reacted to the
! "relationship between one or more subjects", and specifically the
! "one" alternative, since having a relationship between one subject
! sounds kind of odd.

That is the point. The debate narrows down into the association's semantics:
a typed set VS a relationship.