[Haskell-cafe] Top-level state debate on the wiki

[Keean Schupke]

With regards to the following...

"There are cases in which this parameterization costs convenience and 
gains nothing -- for example, the standard library function (randomIO :: 
Random a => IO a) cannot be implemented in Haskell without the 
unsafePerformIO hack, yet there's nothing semantically objectionable 
about it. Worse, we may weaken type checking by the translation, while 
gaining nothing: for example, the interface below cannot provide static 
guarantees like a == b <=> fromAtom a == fromAtom b if toAtom is 
parameterized by an "atom space"."

Can we not make this guarantee with multiple Atom-spaces if we use the 
same local universal quantification trick used on the ST monad? IE add a 
type parameter 's' to the type of Atom?

Keean.