Re: Clean Object Class Design -- Circle/Ellipse

Martijn Meijering wrote:
> hidders_at_win.tue.nl (Jan Hidders) wrote in <9klpmg$3dg$1_at_news.tue.nl>:
> >If you take Date's vision on OO and use a simply type system (no
> >parameterized types or types with quantifiers of types, et cetera) then
> >this is a well-understood problem. You can find the relevant literature
> >under 'multimethods' and 'subtyping' with Google.
> >
> >To summarize:
> [snip]
>
> Thanks for the summary. The difference between this and what I was thinking
> of seems to be that there are no axioms to be satisfied here.

Indirectly they are. In my summary I presumed you know the semantics of the functions, and these might be defined by the axioms, for example.

> Does Date say
> that a decent OO type system should allow you to model any subtype of the
> kind you've just described as an inheritance relationship? I'm not sure it
> matters, but does he make any statements concerning axioms?

I don't think Date has ever been very explicit about that. The nearest thing is probably on:

  http://www.dbpd.com/vault/9902/date9902.shtml

His requirement seem to be only syntactical in the sense that he speaks about the "same operator" where it is not clear if that means just an operator with the same name or with the same semantics.

-- 
  Jan Hidders