Mon Aug 1 17:23:19 CEST 2011


I think I get the big picture, but there's some magic for partial eval
I don't understand yet.  One remark was intriguing:

  ".. to squeeze more invariants out of a type system as simple as
  Hindley-Milner, we shift the burden of representation and
  computation from consumers to producers .."

This is a bit similar to the "printing is simpler than parsing" mantra.

So what's next?  Fix the "let" duplication problem and start using it;
it looks like it's ready.

[1] http://www.cs.rutgers.edu/~ccshan/tagless/aplas.pdf