> > Beyond cycle counting, I see op< as more primitive than op<=: one
> > establishes order, while the other allows for equivalence.

> I see what you mean, I guess,
> but I don't see why this "primitiveness" is a good thing.

I think of op< as the fundamental ordering relation: "Does x precede y?"
Looks semantically cleaner to me to specify SORT's behavior in terms of
this primitive rather than the hybrid op<= ("Does x precede or equal y?"
or "Does y not precede x?").  It's essentially part of the type
declaration for the domain of the SORT.

> Does it make sense to sort sets that are not partially ordered?

Oh!  That must be where you and Olin are coming from.  You see the SORT
predicate as a partial order on the domain of the SORT, and so op<= as
the fundamental relation.