Sun May 1 15:01:04 EDT 2011

serial vs parallel `let'

I wonder if it's actually a good idea to have no parallel `let' form
at all.  I did this to make the VM's continuations simpler
(essentially only 1 since `begin' is the same as single-clause `let'
and a dummy var).

It seems this needs a separate register:
  - current environment
  - environment under construction

So which is simpler?

Moving towards a bytecode VM it seems that the most important
simplification is to have only a single continuation type, such that
it doesn't need tagging and can be a machine code jump target.

It seems to me that a linear environment is probably simpler to later
unify with a stack language.