[sc34wg3] Association items
Sun, 12 Jun 2005 23:40:45 -0400
* Murray Altheim
! [...] It wasn't so much
! that associations were "meant to replace" facets as it was (to my
! memory) that associations are able to model what facets were meant
! to provide. That is, in the simplified model of "facets" as defined
! (minimally) in ISO 13250:2000.
It is a nice way to put it.
! If you mean modify the XTM syntax to provide for them, I must agree
! with the above-stated sentiments that there's no need. From the simple
! facets-as-properties perspective, you can do anything ISO 13250 facets
! can do with the current syntax.
I never had a real need for them, as I was quite fine with stretching
occurrences to represent properties, as most of us obviously have done.
! If you mean more of the complexities
! of a different definition of facets, as perhaps in Faceted Classification,
! you can do any of this using the current syntax and a PSI set. I'm doing
! that now in Ceryle, for example (and in fact use it quite extensively as
! both part of my base ontology and functionally within the software too).
Agreed. I am very interested to see the results of your work.
Actually I have been doing faceted classifications with topic maps since
early 2000. 
I have advanced much further now, of course.
! > On the other hand, even "monadic" associations (re: Murray's posts on
! > other sub-thread) have a few topics related - if you count association
! > and scope.
! As I mentioned in one of my previous posts, this is determined by
! whether you want to bring the things in the model you're discussing
! up to the full status of "entity" (Topic), or if they should be
! modeled as properties (Topic characteristics). Association type
! and scope are not full entities (Topics) in XTM, though they do
! potentially link to them, as you say.
Not sure what you mean by "not full entities (Topics)".
! These are, however, both
! optional in XTM 1.0. I realize that Lars Marius has argued that
! association type (along with other things) should not be optional
! in XTM, whereas I would suggest that XTM be left alone and push
! this kind of thing into a proper constraint language operating at
! a higher schema level -- the XTM DTD is really only meant to
! express syntax-level constraints.
* Lars Marius Garshol
! What we agreed in Amsterdam (this year) was to leave XTM alone, but to
! require an association type in TMDM. However, TMDM will define a PSI
! for untyped associations that is inserted when no explicit type is
! given. Thus we avoid the ugliness of association.[type] sometimes
! being null, yet we keep backwards compatibility.
This sounds great! But wasn't it already here:
Also, unless things have changed, scope is only syntactically optional: if
omitted it was assumed to be the "unconstrained" scope.