TODO

Sadly, all of the below is necessary to even think about running
arc.arc. I might kludge in some things just so I can see arc.arc run,
which would do a lot for my motivation to work on the project.

* Proper lexical scoping
  If possible, we'd like to push this to Python instead of building
  scope chains ourselves like PyScheme's environment.py. The current
  implementation's semantics are very likely to be bogus.

  Python 3's nonlocal may help here, but I don't want to move to
  Python 3. We'll lose Unladen Swallow. OTOH, it looks impossible to
  hack nonlocal support together in a reasonable way in Python itself
  (changeset implementing nonlocal is at
  http://svn.python.org/view?view=rev&revision=53986). With nonlocal,
  I think we can push Scheme/Arc lexical scoping down to Python using
  Arc's trick of keeping a stack of names, although we'll have to
  distinguish the local environment.

* Quasiquote support
  Not implemented at all. Shouldn't be too bad, I think Arc has a full
  implementation.

* Symbol-escape support in the reader
  The | character.

* Simple expansion in the reader
  Support Arc's "~", "'", "`", ",", ",@" operators.

* More complicated expansion in the reader
  Support Arc's ":" operator. 

* Macro expansion.


Nice to have:

* Expose more of Python in a sane way.

* Decide what, if anything, to do about tail calls.
  Python doesn't optimize them.