[MLton] HOL's Moscow ML implementation, and pushing MLton to emulate it

[Daniel C. Wang]

Oh, I must have grossly misread your code.. originally.. I see what you are 
getting at now.

My example code fixed a universal type, and had the host code deal with the 
mismatch between native and universal type. What you are suggesting is that 
we do the reverse. Fix a representation that's good for the host and make 
the hosted code deal with the rep mismatch. I think that this makes perfect 
sense.

One question comes up. Does this mean the interpter has to be rewritten to 
acess its universal type via some abstract interface?