[om-list] relationships in OM

[Mark Butler]

Luke Call wrote:
> 
> In thinking about modeling relationships in OM (and in working on the
> use cases to propose to the group), I want to suggest for your feedback
> that a class is really just a query. Here's why. There are many types of
> classes, say "picture frame", and "aluminum object" and "border" which
> may not fit into neat hierarchies, even with multiple inheritance. But a
> query can resolve any grouping if the criteria are on data stored in the
> system. And a query can be stored with a name and other attributed if
> needed, just like a class.

For a large number of synthetic classes that approach works just fine.  In
addition, although somewhat inefficient, we can certainly test for class
membership against a class membership definition predicate or query.  The only
problem is getting the query predicate precise enough.

The difference between the query mode and the standard mode of class
definition, is that in the query mode objects automatically become members of
the class by meeting all the terms of the query, i.e. the query defines the
class.  In the standard mode, the fact that an object has identical or
compliant attributes to the class does not make it a member of the class
without someone declaring it to be so.