emacs 
-----

* set the compile command to "tail -f cat.log" to use compilation-mode
  for easy error navigation.


naming conventions
------------------

* a 'word' is a single (named) code entity in a compositional
  language. the strict meaning in CAT is the element of a program =
  list of words. in PURRR/18, a word refers to a code address, in
  PURRR a word refers to a primitive address (execution token), and a
  program is an array of words (run token).

* a procedure (or function) is a scheme code entity.

* words prefixed with '_' (underscore) refer to double wordlength
  operations. each underscore doubles the wordlength.

* '>' and '<' in names indicate linear data motion (not copying!). ex: '>r'

* '->' and '<-' in names indicate data conversion.

* '>>' and '<<' are named SHIFT RIGHT/LEFT and indicate bit shifts.

* '>>>' and '<<<' shift n bits (word n -- shifted)

* words/procedures prefixed with '~' (tilde) implement the core part
  of something, missing some (shared) preprocessing step.

* a->xxx and f->yyy set the current RAM and FLASH ROM pointers to some
  object.

strings
-------

i am using 'Pascal strings', instead of NULL terminated 'C
strings'. pro/con :

+ can contain arbitrary bytes (no speciall NULL byte)
+ only require a single pass when length needs to be known

- limited size

on a small chip like a pic buffers are usually small, so it makes
sense to design code so the limited size is not a problem. this will
greatly simplify code.

strings stored in flash memory are accessed using the 'f' register
(TBLPTR). this register can be understood as the 'current flash object
register'.

suppose f points to a string, performing an operation on a string is
then simply:

: xxx-string   @f++ for  
  	       	    	 @f++ xxx  
  	       	    next ;

where 'xxx' consumes one byte of the string.