On Mon, Dec 03, 2007 at 03:32:52PM +0200, Yitzchak Gale wrote:
Henning Thielemann wrote:
...it is not sensible to have different instances for the same type and class, because they will collide sooner or later.
True. That is why libraries should not define an instance at all, unless they are quite certain that it is by far the most important instance that anyone will ever want to use.
I would draw the opposite conclusion from the same data: if a sensible instance can be identified, it should accompany either the class or the type constructor. If people define orphan instances, they will eventually collide, even they are identical. (There are a few orphans in Control.Monad.Instances, but that is required to preserve compatibility with Haskell 98.) But in the case at issue, the proposed Bounded instance is counter-intuitive because the underlying Ord instance is. That Ord instance is an arbitrary choice that is accepted because it allows IntSets to be used as search keys; it makes no sense on its own. A Bounded instance would attach unwarranted significance to this arbitrary ordering. It might be different if there were a use in sight for the Bounded instance.