Speaking as someone who went from Haskell to ML, I don't see the loss of type classes as a terribly big deal. The implicitness of type classes is useful “in the

Speaking as someone who went from Haskell to ML, I don't see the loss of type classes as a terribly big deal. The implicitness of type classes is useful “in the small”, but the global benefits of a real module system in terms of type abstraction and separation of concerns outweigh the local benefits of the occasional shorter definition using type classes. In addition, ML modules have structural signatures, which makes it easier to retrofit existing modules to new hierarchies. For comparison, try making Haskell's existing Monoid class a subclass of Semigroup, without recompiling any code. :-p