[sc34wg3] Association items

Thu, 09 Jun 2005 10:08:31 +0200

* Patrick Durusau
| 
| I am trying to model the various parts of the TMDM as assertions and
| in the course of re-reading the TMDM, I ran across the following at
| 5.7 Association Items:
| 
| "An association is a representation of a relationship between one or
| more subjects."

It's a bit strange (to me, anyway :) that you should choose this
particular bit of prose as the departure point for the discussion.
The important part is this:

  [roles]: A non-empty set of association role items. The association
  roles for all the topics that participate in this relationship.

Ie: associations must have 1 or more roles.

| Note 41 (from ISO 13250) says: "Thus, the containing assoc element
| can assert that a topic has one or more specific relationships to
| itself."

That's not relevant to unary associations. The context is the
preceding sentence:

  Within a single association link, more than one assocrl element may
  reference the same topic, in which case the topic plays multiple
  roles in the association.

So that would be more like

  is-in-love-with(narcissus : lover, narcissus : loved)
    /* unfortunately, this is not a symmetric relation */

which is what the NOTE is talking about.

| Although it is represented in syntax as a "unary" association,
| shouldn't it be modeled as an association with two roles, etc.?

Why would we do that? What would be the type of the other role, what
would be the player? IMHO, unary associations are very useful for
expressing binary conditions like

  is-consistent(lmg : consistent)

and I can't really think of an implied second role that would make any
sense.

| This is not the only place where the issue comes up, as topic name,
| for example, is mentioned as a specialized kind of occurrence (TMDM,
| page 11) and it is later noted that an occurrence is a specialized
| form of an association (TMDM, page 13).

Well, you can't have a "unary name" or "unary occurrence", so this is
not a problem. If you have a name or an occurrence they both *must*
have a topic they are attached to.

| NOTE: I am NOT suggesting any change in syntax, processing, etc.,
| but am asking if the "unary" association in syntax should be modeled
| as though it were an association with two (or more) roles, role
| players, etc. 

Well, this would be a change in processing, but if it were the right
thing to do we might change it anyway. So far I've seen nothing to
indicate that it is the right thing to do.

| Some parts of which are implied, for example with topic name if it
| is considered ultimately to be a form of an association (I take it
| the roles are implied even though the role players are known).

Yes, you could think of them as being implied, but in practice that's
not a consideration as in the structure there aren't actually any
roles, and so what their types might be is purely an academic
question.

You mean that it would be equivalent to

  is-inquisitive(jan : person, is-inquisitive : characteristic)?

That seems like a bizarre interpretation to me. Why would we do this?

-- 
Lars Marius Garshol, Ontopian         <URL: http://www.ontopia.net >
GSM: +47 98 21 55 50                  <URL: http://www.garshol.priv.no >