Supertypes

What supertypes are

If a type S is a supertype of type T, the basic attributes and methods of type S can also be applied to objects of type T; and the semantics of objects in type T is consistent with that of objects in type S. We say that type T is a subtype of type S, and that objects of type T (and of type S) conform to type S.

Subtyping is transitive. If type R is a subtype of type S, and type S is a subtype of type T, then type R is a subtype of type T.

A type may have multiple supertypes.

How supertypes are defined

Many subtyping relations exist implicitly, but TOP will not consider one type to be a subtype of another type unless the supertype relation is explicitly declared in the description of the subtype, or unless there is a chain of such declarations leading from the subtype to the supertype.

Subtype declarations essentially make guarantees about the behavior of the subtype. They may require some sort of explanation or abstraction function showing how the subtype conforms to the supertype. But new supertype declarations should not actually introduce any new semantic constraints that weren't in the original definition of the subtype. In order to prevent such constraints from being spuriously introduced, TOP servers will often constrain who is allowed to make new supertype declarations.

Supertypes and substitutability (some formal notes)

Those familiar with object type theory may recognize that the notion of subtyping above is basically equivalent to some formal notions of "substitutability". One should, however, note the following differences between TOP's subtyping and the subtyping that exists in some object-oriented systems that use substitutability:
spok@cs.cmu.edu (Last updated 11-Jan-95)