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

[Stephen Weeks]

> > Sort of.  But he imposes a *severe* restriction on what host
> > representations are allowed, that in particular rules out many
> > representations that MLton uses.
> 
> Hmmm perhaps here is where the point of confusion is. I don't actually think 
> that is the case. There are restrictions on the representation of the 
> universal type, but not on the underlying native rep.

But one must be able to coerce in constant time, so those restrictions
on the universal type impact the underlying rep as well.

> Can you give me an example of what exactly you mean in the context of a 
> list? What MLton opts are you thinking about?

Sure.  MLton represents a Real64.real list using 12 bytes per list
element, while it represents an Int32.int list using 8 bytes per list
element.