Tue Nov 16 08:27:28 EST 2010

Summary of possible directions

In order of importance

1 Fix module system: proper units and eliminate "load".  The USB
  driver can be the pull for this.

2 Keep the "eager macros" partial evaluation strategy, but augment it
  with a semantics of the low level machine language used.

3 Build the compiler on top of a more abstract control flow graph.
  Currently the way compilation state is maintained feels a bit raw..

4 Build a theory for the I :: m -> (t -> t) towering in Haskell.