[MLton] Re: applicative cfg

[Norman Ramsey]

 > One question I had was whether there were contexts other than the UW 
 > optimization framework where you found that the property list approach of 
 > MLton seemed more natural than the unique finite map approach of GHC.

I think the real question may be 'threading or not threading'.
If we had a mutable cell containing a finite map, for example, it
would be just fine.  Our property interface hides this distinction.
I guess we should clarify in the paper :-)

 > You could improve the parallel structure at the beginning of Section 2.3:
 > 
 >   * The glorious Glasgow Haskell Compiler (GHC) uses ...
 >   * The   MLton Standard ML compiler uses ...
 >         ^
 >         magnificent

The 'glory' is an insider's reference to Will Partain's comments in
the GHC source code as well as the official documentation.  Until a
Google search for 'magnificent mlton' does as well as a search for
'glorious glasgow', I'm afraid MLton will have to remain unmodified :-)


Norman