Reference types

Ashley Yakeley [email protected]
Thu, 7 Feb 2002 02:32:11 -0800


At 2002-02-07 02:09, I wrote:

>The kind of generalisation you are proposing is, in my opinion, best done 
>explicitly by Haskell. Primitive functions and types should be as simple, 
>concrete and primitive as possible. Let Haskell do the clever 
>generalisation stuff.

As a rule, I'm opposed to any generalisation in the standard libraries 
over IO and (ST s) that cannot be made to work for user-defined monads. 
People will use them in their code, and those monads will become 
"privileged". I admit I have a stake in this, my JVM-Bridge makes 
extensive use of lifted monads, and so I'd like porting code between 
monads to be as easy as possible. The way forward for this is classes and 
types in the standard libraries that generalise over any monad which has 
the necessary properties.

-- 
Ashley Yakeley, Seattle WA