Fri Jan 2 18:51:30 EST 2015


- Dynamic Erlang.  Needs heap and GC, proper datastructures, ...  This
  won't work on PIC but might be feasible on ARM.

- Static Erlang.  (Static tasks, stacks, no heap, flat messages).
  Does this produce a usable primitive set for programming?

- Full actor language, partually evaluated to synchronous state
  machine.  E.g. don't speficy "compilable subset".  See this as a
  compile time test (works / wont work, here's why).

- More traditional Forth multitasking / coroutines.

Which one is the more interesting approach?  I'd say the full actor
model.  There is an implementation of this in Scheme but since it
might be necessary to take the scheduler into the abstract
interpretation loop let's write it from scratch.