Re: Typing and semantics [dynamic vs. static typing]

[John Clements <address@hidden>]

On Dec 15, 2003, at 5:15 PM, Kevin S. Millikin wrote:

>> No, wait a minute. I'm not excluding semantics that use type
>> information, just those that use the result of the type proof.
>
> But it's not fair to Haskell to not allow it to use the results of type
> inference later on, when an explicitly typed language *is* allowed to
> use the programmer supplied type annotations (i.e., the results of
> manual type inference).
>
> Is the real beef with type inference?  Certainly, type inference is
> hard to understand if the programmer cannot be made to understand the
> type inference algorithm.  Haskell programmers seem to learn to cope.
>  They are using a lazy functional language, after all.  Nothing else
> makes sense anyway!

My beef is _not_ with type inference.  I claim (as I have in a response 
to Ken) that ML has a deterministic operational semantics.

Cribbing again from my response to Ken, I think the core of the debate 
has to do with typeclasses as a program abstraction mechanism that is 
inherently non-local.

Your point about laziness is well-taken.

john