Text 20146, 84 rader
Skriven 2011-06-26 04:56:40 av Bob Ackley (1:300/3)
  Kommentar till text 20121 av Gary Perkins (3817.fido_news)
Ärende: Character sets
======================
Replying to a message of Gary Perkins to Michiel van der Vlist:

 GP>   Re: Character sets
 GP>   By: Michiel van der Vlist to Gary Perkins on Sat Jun 25 2011
 GP> 03:36 pm

 >> Good point. When memoy was measured in kB rather tham GB, doubling
 >> the required memory by reserving 16 bits for a character was not an
 >> option.

 GP> Probably more like just bytes..lol

 >> The available 256 positions were not used  all that effieciently
 >> though. 32 positions for control characters seems a bit overdone in
 >> hindsight. Not more than a handfull were ever used. There was also
 >> the somewhat debatable choice of reserving the most significant bit
 >> as the parity bit. In retrospect, that was not the smartest thing
 >> to do. Well, hindsight is always 20/20.

 GP> Original 128, you mean.  I haven't looked it up yet, but the
 GP> original control characters were probably used a lot more in the
 GP> older days for terminal and data control.  Some are probably only
 GP> kept for legacy purposes now.  Others still have some use.

 >> OTOH, it is no use denying that ASCII was designed by Americans with
 >> little or no consideration for the needs of others. The 'A' in
 >> ASCII sya a lot. Also, the fact that there is a Dollar sign, and no
 >> Pound sign, or a "general currency" sign, says a lot.

 GP> I always thought other areas had their own code tables.

Actually, the reason for ASCII had little or nothing to do with various
countries'
character sets.

Way back in the day, each computer manufacturer (there were several besides
IBM)
had its own proprietary character set, and all were incompatible with each
other.
ASCII was developed so information could be exchanged between say, Honeywell
and Univac machines.  At the time, computer manufacturers used 6-bit, 7-bit and
8-bit characters, ASCII is a happy medium at 7 bits.

The old Honeywell 6000 mainframes I worked on used a 6-bit character set, max
64
characters, 00-77.  It was an OCTAL (base 8) machine rather than hexadecimal
(base 16),
and used 36-bit words rather than the 32-bits most everybody else used.  Memory
was
addressable by word rather than by byte, and there was a hardware limit on how
much 
memory could be addressed - 256K words (in two refrigerator-sized cabinets of
128K
each)  That's the approximate equivalent of 1 megabyte of RAM; state of the art
- 1970.

FWIW, the original 'control codes,' represented today by 00-0F in ASCII, were
used
for control of teletypes.  Letter shift, figs shift, bell, CR and LF were
pretty much all
that a Model 28 teletype used (I learned to type on one of those, doing data
entry - jeez,
*40+* years ago <sigh>).

Not that it matters a hill of beans today, but GE, Sperry, Burroughs,
Honeywell, RCA,
Remington Rand, NCR and Control Data Corp. all made mainframes back in the
1960s.
So did Xerox and Western Electric (the WE 3B2 is the machine that Unix was
originally
developed on and for); note that Western Electric only sold its equipment to
AT&T, its
corporate parent.  GE and RCA both sold their computer divisions to Honeywell.
Sperry bought Univac from Remington Rand and then merged with Burroughs to form
Unisys.  Honeywell sold its computer division to Bull, which is a French
company.
Dunno what happened to CDC, FWIW Cray is/was an offshoot of CDC.  AFAIK NCR
finally gave up on their computer division after AT&T bought the company, their
Criterion 8400 was a nice box - I worked on one for a couple of years.

Also gone by the wayside:  Data General, DEC, Wang... <sigh>

--- FleetStreet 1.19+
 * Origin: Bob's Boneyard, Emerson, Iowa (1:300/3)