My 22nd published paper, "Do It by the Numbers" [1], was
triggered by the old story of the comedian's club laughing when
someone said a number, because they all knew the joke so numbered
in their comedian's handbook.
The scheme used a dictionary of common words, ordered not
alphabetically but by usage frequency. These words were given
binary code equivalents of variable length, but self-terminating
by a rule. Assignment was in binary order by length in number of
bits, until the combinations of that length were exhausted, and
the length bumped by 1. E.g., the 4 most frequent words were of
2 bits and terminator; the next 8 most frequent words were of 3
bits and terminator; etc.
The transmission method used a similarly equipped computer at
each end. One encoding and sending, the other receiving and
decoding. Moreover, the dictionary was not static. As new words
entered, and previous frequencies changed, their encoded
assignments changed for both computers. This was useful for
cryptographic purposes.
The article had caused no stir in itself. But then IBM decided it
would make a nice publicity splash. It appeared this way in the
New York Herald Tribune:
Second Section, Page 1 -- 1960 July 08
|
That set off a lot of correspondence. David Kahn mentioned it in
his classic book "The Code Breakers". He said later that even if
IBM had not made it a commercial success, it was nevertheless
probably very resistant to being broken.
Someone at Ferranti in the UK asked if they could use the
method for overseas telegraphic transmission. I checked
with IBM, who then considered it only for publicity, not for
an actual product, and replied "YES" to the querier, one Conway
Berners-Lee.
Meanwhile, my group at IBM was making a program product of it,
via two great programmers, Mark Halpern and Bob Brill [2]; in the
UK, Berners-Lee also set about doing having it done.
|