Tue Aug 5 09:32:01 CEST 2008

ANS Forth

It seems that standardizing is an essential part to get to some
adoption. Basicly, nobody cares about nonstandard Forths: people write
their own. Makes sense really.

So, let's bring the PIC18 Forth to standard. The goal is to do it as
convenient as possible, without loosing too much time on
optimization. Some ideas I gathered before:

  * Data doubler: Add a layer that performs just primitive data size

  * Unified address space. Map part of RAM into the namespace.

  * Interpreter. It seems a good idea to stick with a native Forth,
    and write a dispatching interpreter on top of this. That way all
    primitives can be re-used.

It's simplest to first make the data doubler, so words written in the
doubled language are usable in the unit language. On top of this
memory access words can be written, which then can support a trivial
interpreter loop.

Alternatively, I can implement Taygeta's primitives, and optimize them.