the MPTC Dilemma (please solve)

[Martin Sulzmann]

Manuel M T Chakravarty writes:
 > > Another thing, here's an excerpt of the current summary of the MPTC
 > > dilemma from
 > > http://haskell.galois.com/cgi-bin/haskell-prime/trac.cgi/wiki
 > >
 > > 
 > > Multi Parameter Type Classes Dilemma ¦
 > > 
 > >     Options for solving the dilemma ¦   
 > >         2. Put AssociatedTypes on the fast-track for sainthood
 > >
 > > Associated Types ¦
 > > 
 > >    Cons ¦   
 > >     * Only a prototype implementation so far 
 > > 
 > > This is a *inaccurate* and highly *misleading* summary.
 > 
 > I didn't write that wiki page plus you are citing individual sentences
 > out of a larger text.

This wasn't meant as a "personal attack". If you (or whoever wrote)
feel offended, I'm sorry. 


Fact is:
(risking that I repeat myself here)

 - When it comes to typing,  FDs strictly subsume ATs.
   (FDs = as defined in the FD-CHR paper) 
   FYI, the next Chameleon release will supports FDs and ATs,
   ATs are internally represented via FD-CHRs.
 - All the challenging inference problems encountered for FDs
   also apply to ATs (see my previous emails)
 - FD-CHRs describe improvement *and* instance, superclass rules.
   It's really the subtle interaction between improvement and
   instances that causes the problem.
   
I also would like to point that none of the FD-CHR
results are "recent". The FD-CHR paper was published in ESOP'04,
that was two years ago. Since April 2004, there's an extended
version available as a TR. Correct, a revised version will somewhen
soon appear as a JFP paper.

Martin