[PRL] Type-directed partial evaluation

[John Clements]

On Jan 8, 2004, at 4:15 PM, Carl Eastlund wrote:

> But it does suffer from that problem.  "S" needs to generate a "C", 
> which
> it can only do using its first paramter (A -> B -> C), so it also 
> needs an
> A and a B.  Which it can only get from its other two arguments, and the
> constraints stay the same.
>
> Whether this actually results in a problem in Dave's code and/or 
> theory, I
> don't know.
>
> --Carl

yes, well, what about '(A -> A -> A)? Could be (lambda (x) (lambda (y) 
x)) or (lambda (x) (lambda (y) y)).  So that's not hard.

john