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

[Matthew Fluet]

> Stephen Weeks wrote:
> {stuff deleted}
> > To spell out a little more detail -- this does not involve rewriting
> > all of MLton!  
> {stuff deleted}
> 
> BTW this sounds awfully like an approach that would end up giving us 
> separate compilation!. i.e. compile two programs with MLTon link them 
> together and have them interoperate via a universal data rep...

One had to defunctorize the program before monomorphising, and one still 
needs to defunctionalize after monomorphising.