The "EEL Compressed Source" file format
	---------------------------------------

EEL Compressed Source files are EEL source files stripped of
all comments and formatting information. They are tokenized
using an external dictionary, which contains all keywords of
EEL and extensions, and any project names. Immediate values
are encoded using an extended "variable length number" BCD
format.

* EEL Compressed Source files should have the extension
  ".eec" instead of the normal ".eel", for reliable
  identification, and to avoid accidentally overwriting
  source files with compressed source files. (That would
  be nasty, as the compression is destructive WRT comments,
  symbol names and code formatting...)

* The files are based on 4 bit words. Wider words (8 or 16
  bits) are *not* aligned to 8 or 16 bits, but start
  directly at the current 4 bit word position.

* There are two modes: XASCII and XBCD. Decoding of a file
  starts in the XASCII mode. XASCII codes are 8 bits, and
  XBCD codes are 4 bits.

* The file starts with an 8 bit version code for the EEL
  dictionary to use. Then follows a list of 4 byte EEL
  extension IDs, terminated by a 0 byte. The EEL extension
  IDs are of the form "XXXv", where "XXX" is an acronym
  identifying the extension, and "v" is the version of the
  extension dictionary to use.
  
  Any extension ID that doesn't math any known EEL
  extension is assumed to be a Local Dictionary name.
  ".eed" will be appended to the full 4 byte string (no
  version code), and then EEL will try to find a file by
  that name and load it as a Local Dictionary.

* A Local Dictionary file consists of a 32 bit little
  endian integer that holds the number of tokens needed
  by the dictionary, *optionally* followed by a list of
  null terminated ASCII strings, holding the human
  readable names of all tokens. (The token name list is
  useful when mixing compressed source with ASCII source
  during development.)

* Token values are allocated from 0 and up as
  dictionaries are loaded. This is why there are
  extension version numbers in the dictionary lists -
  and it should also make it obvious that EEL extensions
  must only *add* to their dictionaries if compatibility
  with old compressed sources is to be maintained.

* XASCII code:
	0-14	XBCD code + enter XBCD mode

	15	Local token dictionary follows
	16	8 bit character follows
	17	Null terminated 8 bit quoted string
	18	Quoted 8 bit Pascal string follows
	19	Quoted 8 bit Pascal string continued
	20	<unused>
	21	<unused>
	22	<unused>
	23	<unused>
	24	<unused>
	25	<unused>
	26	<unused>
	27	<unused>
	28	<unused>
	29	<unused>
	30	<unused>
	31	<unused>

	32-127	Standard 7 bit ASCII (copied as is)

	128-191	Short Tokens
	192-254	Extended Tokens (8 extra bits follow)
	255	Wide Tokens (16 extra bits follow)

* XBCD code:
	0-9	Decimal numbers 0-9
	10	Decimal point (.)
	11	Unary minus (-)
	12	Argument separator/white space
	13	Statement terminator (;)
	14	Single XASCII character follows
	15	Switch to XASCII mode