On Thu, Oct 30, 2003 at 09:05:04AM -0800, Tom Lord wrote:
>
> Perhaps in all cases, the optional equivalence predicate to a
> constructor is mostly just a hint.
>
> If I make a hash table using EQV? but then modify it with:
>
> 	(dictionary-set! d k v)		; set using equal?
>
> instead of
>
> 	(dictionary-setv! d k v)	; set using eqv?
>

>
> then perhaps the results are simply undefined.   The non-generic
> interface to (normal) association lists works that way, for example.
> I've certainly used hash-table libraries that work that way.

But why build in so much semantic uncertainty when the utility of
changing equivalence in flight is limited to what I imagine are some
pretty unusual cases?

	Scott