On Sun, Sep 29, 2013 at 10:33 PM, Simon Peyton-Jones
As Edward says, there is representational equality too, and we want both a constraint and a data type for it. Currently the constraint looks like Coercible a b but and the data type is currently called EqR. Linking them more tightly could be a Good Thing. For example, we could hav a ~~ b for the constraint and a :~~: b for the data type. I do not have strongly feelings; just pointing out the correspondences.
Simon
I think `Coercible` and `coerce` (along with `unsafeCoerce`) are perfect as they are. Not everything has to be an operator. Things with names are self-documenting, which is good. For something that's going to show up very frequently, it makes sense to use an operator, because (a) by showing up very frequently, it becomes common knowledge, and (b) it's shorter. I think this is true for (~). If something shows up only occasionally, I think it makes more sense to use a name, because (a) widespread familiarity can't be assumed, and (b) brevity is less important. I think this is true for `Coercible`. Of course, this doesn't answer the question of what to call the data type. :-) I don't have any great ideas. (Coerce? CanCoerce? IsCoercible?)