On Wed, May 09, 2007 at 03:08:03PM -0700, Andrew Lentvorski wrote:
> For a while now, I've wondered whether Morse code would be a better way 
> to do text messaging than what we have.  Especially since Morse code 
> only takes a single button and did a sane job of Huffman encoding.
> 
...
While Morse code does a good job with a single button, this isn't
exactly the same problem as Huffman encoding.  Huffman encoding
requires that any string of ones and zeros be valid and assumes
that ones cost the same amount of time as zeros.

Morse has different constraints, some of which become apparent
if we look at the two ways people try to treat Morse as binary:

  1.  If you consider dots as zeros and dashes as ones, you need to
  use base 3 instead of binary so you can represent the spaces between
  letters and between words, e.g. "from me" needs to
  be ..-.x.-.x---x--xx--x. , 
  not just ..-..-.-------. , which may be ambiguous depending on context.

  2.  If you consider no carrier (key not pressed) as zero and
  carrier (key pressed) as one, it is binary, but if you send too many
  consecutive zeros or ones the receiver won't know precisely how many
  you intended to send.  Morse solves this problem by requiring strings
  of ones to be only 2 lengths (dots and dashes), and strings of
  zeros to be only 3 lengths (symbol, letter, and word separators).
  A textbook description (1957 Bluejackets' Manual) gives nominal
  lengths, so we can write (approximately)

  0 = symbol separator
  1 = dot
  000 = letter separator
  111 = dash
  0000000 = word separator

  Now "from me"
  or ..-.x.-.x---x--xx--x.
  becomes 1010111010001011101000111011101110001110111000000011101110001

Despite the length of this string, the empirical evidence is that
Morse is fairly close to optimal.  (You can quibble about things like
O, C, and Y having overly long encodings and M, W, X, and Z too
short.)  Because dashes are longer than dots, they are avoided to some
extent for common letters.  Morse is somewhat less compact compared
to other encodings in
contexts where dashes are just as short as dots, e.g. in puzzle
ciphers like "Fractionated Morse" that use Morse as a first step with
the base 3 representation (like ..-.x.-.x---x--xx--x.), or in
communication if you provide separate dot and dash buttons and
transmit them in some way that makes them have equal length.

I apologize if my lack of experience with Morse has led
me astray.

Stewart Strait


-- 
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to