Is a byte always 8 bits ?
Richard Flamsholt S0rensen
This may not exactly be considered folklore, but still, a.f.c seems
like The Right Place to bring this up.
I've always thought, that a byte was defined to be 8 bits. However,
a couple of years ago I read somewhere that "it *usually* contains 8
bits, but this is not the definition of a byte".
Confusion, confusion - after that, I did some research and
discovered, that distinquished authors (like Tanenbaum) (sp? <- how
embarrasing!) tend to disagree on the matter, so now I turn to you
net.people hoping to set the record straight.
Richard
--
/Richard Flamsholt
ric...@iesd.auc.dk
Mark Crispin
> I've always thought, that a byte was defined to be 8 bits. However,
>a couple of years ago I read somewhere that "it *usually* contains 8
>bits, but this is not the definition of a byte".
On the PDP-10, a byte could be any consecutive group of from 1 to 36
bits, positioned anywhere within a 36 bit word. Bytes were referenced
by one of three kinds of pointers:
. a local pointer, within the same "section" (approximately 1
megabyte) of memory, with a 6-bit field indicating position, a
6-bit field indicating size, and a 23-bit field indicating
indirect (1 bit), index register (4 bits), and base (18 bits).
This was the only type of pointer on processor models limited to
the traditional 18-bit (36-bit words, so ~1 megabyte) address
space.
. a global pointer, which could access anywhere in memory. The first
word had a 6-bit position, 6-bit size, and 1 bit (always zero in a
local pointer) set to indicate global pointer. The rest of this
word was unused. The second word had 1 bit indirect, 4-bit index,
and 30-bits address.
. a special global pointer, which had a 6-bit code and a 30-bit
address. The code was various values for the common positions of
6, 7, 8, 9, and (I think) 18 bit bytes.
There were also special instructions to handle half-words (18 bits),
full-words (36 bits), and double-words (72 bits).
_____ | ____ ___|___ /__ Mark ("Gaijin") Crispin "Gaijin! Gaijin!"
_|_|_ -|- || __|__ / / R90/6 pilot, DoD #0105 "Gaijin ha doko?"
|_|_|_| |\-++- |===| / / Atheist & Proud "Niichan ha gaijin."
--|-- /| |||| |___| /\ (206) 842-2385/543-5762 "Chigau. Omae ha gaijin."
/|\ | |/\| _______ / \ FAX: (206) 543-3909 "Iie, boku ha nihonjin."
/ | \ | |__| / \ / \M...@CAC.Washington.EDU "Souka. Yappari gaijin!"
Hee, dakedo UNIX nanka wo tsukatte, umaku ikanaku temo shiranai yo.
Lance Taylorism Smith
> I've always thought, that a byte was defined to be 8 bits. However,
>a couple of years ago I read somewhere that "it *usually* contains 8
>bits, but this is not the definition of a byte".
Just to add the historical perspective to this. The 8-bit byte seems to have
been created within the IBM 7030 development (better known as Project Stretch).
This would put it's origin around 1959-1961. It could be argued that a "true"
byte has 8 bits (just because the original byte did), but other machines have
defined byte differently.
The byte probably came into wide scale use with the IBM System/360. Since there
were so many of this family around, I'm guessing that's what makes us think in
terms of a 8-bit byte. Anyone who knows the system better than I do want to take
a shot at this?
--
Lance "Simon" Smith and his dancing bear
Oh, who would think a boy and bear could be well accepted everywhere...
REPLYTONET: lsm...@cs.umn.edu
Send monetary units to: PO Box 13345, Dinkytown Station, Minneapolis, MN 55414
Tim W Smith
GI 1600 (the microprocessor used in the Mattel Intellivision):
5 bits: nickle
8 bits byte
10 bits decle
16 bits word
This processor had 10 bit wide ROM and 16 bit wide RAM, hence
the need for such a strange combination of sizes.
Tim Smith
ps: the processor used 10 bit ROM because all instructions
were 10 bits wide. According to rumor, GI claimed that they
had figured out that they only needed 10 bits per instruction,
so they designed the CPU to use 10 bit ROM. This was supposed
to be more efficient than other CPUs because, presumably, the
number of bits you would really need would not match the 8 or
16 bit ROMs of other systems, leading to waste. Your ROM space
would be used more efficiently with the 1600, leading to a need
for less ROM, and thus leading to a less expensive system.
Of couse, no one but GI *made* 10 bit ROM, so there was nothing
to drive the price down, so you ended up paying more for the
ROM than if you had used some other processor with "wasteful"
8 or 16 bit ROM!
Sylvia Dutcher
The PDP-6/10 had a wonderful LDB (load byte) in which you specified the
byte size and address. (As well as an even more wonderful Increment
and Load Byte ILDB, which did the auto-add of the partial addresses
for you). E.g. ILDB AC,[440700,,data] would load the first 7 bit byte
from address 'data.' (Note that 44 is the word length in octal bits).
Mark Zellers
>
> This may not exactly be considered folklore, but still, a.f.c seems
>a couple of years ago I read somewhere that "it *usually* contains 8
>bits, but this is not the definition of a byte".
>
> Confusion, confusion - after that, I did some research and
to and including a word. There was an object called a "byte pointer"
which told you where in the word and how many bits to take. The 10 was
the only machine I ever worked with with this capability. In normal
usage, a byte on the 10 was 7 bits wide (which left one spare bit at the
bottom of every word).
Mark H. Zellers
decwrl.dec.com!batgirl!bwayne!mark
Dave Ihnat
computing arena. Remember, for instance, the 36-bit word on the Univac 1100
series of machines, in which a quarter-word--which filled the position of
a byte in processing--was 9 bits...
Bill Hooper
>GI 1600 (the microprocessor used in the Mattel Intellivision):
>
> 5 bits: nickle
> 8 bits byte
> 10 bits decle
> 16 bits word
>
Bill (yes, i do have better things to do) Hooper
w...@hpncmo.rose.hp.com
Ray Charbonneau
----------------------------------------------------------------------------
Ray Charbonneau | Appropriate remark! - Dave Barry
rcha...@athena.mit.edu | If it wasn't for drinking and driving,
| how would I get home? - Steve Sweeney
----------------------------------------------------------------------------
MIT and the Library Systems Office are glad that I'm posting my opinions
her so that they don't have to listen to them anymore.
----------------------------------------------------------------------------
Paul Ahlgren
GEORGE 3, where the bytes were 6 bits and they normally worked with words
that were 4 bytes ( or 24 bits).
It also did most of its counting in octal mode ( tape numbers etc...)
--
Paul Ahlgren Digital Equipment Corporation
Customer Services Operations, Sweden
Telephone : [46] - (8) - 733 73 56
These are my opinions, and my opinions alone. They should not be
confused with the ones of my employeer.
John Baldwin
rcha...@athena.mit.edu (Ray Charbonneau) writes:
>4 bits is a nybble, 8 bits is a byte, and 32 bits is a dynner.
I guess 64 bits makes a tyummyache.
--
John Baldwin | srchtec!jo...@gatech.edu
| johnb%srchte...@mathcs.emory.edu
Kurt Hirchert
It definitely was not 8 bits!
--
Kurt W. Hirchert hirc...@ncsa.uiuc.edu
National Center for Supercomputing Applications
NRAO Array Operations Center
>Back in the beginning of the 80's I was working on an ICL 2960 machine running
>GEORGE 3, where the bytes were 6 bits and they normally worked with words
>that were 4 bytes ( or 24 bits).
At least some of the Harris machines do (did?) this. One that my husband has
had to transport data from I believe uses 6 bits/byte, and also uses weird
type declarations in Fortran like REAL*6 and INTEGER*3. Truly bizarre.
--
Ruth Milner
Systems Manager NRAO/VLA Socorro NM
rmi...@zia.aoc.nrao.edu
Eric S. Raymond
> In article <1990Dec12....@rice.edu> j...@moray.rice.edu (Shawn Joel Dube) writes:
> >
> >In article <1990Dec12.0...@athena.mit.edu>, rcha...@athena.mit.edu (Ray Charbonneau) writes:
> >
> >|> 4 bits is a nybble, 8 bits is a byte, and 32 bits is a dynner.
> >|>
> >
> >16 bits is a word and 2048 bits is a page.
>
> and 2 bits are a snak!
You forgot 16 bits = `playte' to fill out the y-series. I don't think this
has ever been used as other than a joke, though. These are all in the new
Jargon File.
--
Eric S. Raymond = er...@snark.thyrsus.com (mad mastermind of TMN-Netnews)
Steve Lamont
True. ANSI gets around the ambiguity of what constitutes a "byte" by defining
something called an "octet," which, indeed, is 8 bits.
spl (the p stands
please, now, can we
get on with the
folklore? Anybody got
any good "walking disk
drive" stories?)
--
Steve Lamont, SciViGuy -- 18...@cc.nps.navy.mil -- a guest on network.ucsd.edu
NPS Confuser Center / Code 51 / Naval Postgraduate School / Monterey, CA 93943
What is truth and what is fable, where is Ruth and where is Mabel?
- Director/producer John Amiel, heard on NPR
Frederick Roeber
> In <1990Dec13.1...@sdd.hp.com> Tim McDonough wrote:
>> In article <1990Dec12....@rice.edu> j...@moray.rice.edu (Shawn Joel Dube) writes:
>> >
>> >In article <1990Dec12.0...@athena.mit.edu>, rcha...@athena.mit.edu (Ray Charbonneau) writes:
>> >
>> >|> 4 bits is a nybble, 8 bits is a byte, and 32 bits is a dynner.
>> >|>
>> >And to continue it....
>> >
>> >16 bits is a word and 2048 bits is a page.
>>
>> and 2 bits are a snak!
>
> You forgot 16 bits = `playte' to fill out the y-series. I don't think this
> has ever been used as other than a joke, though. These are all in the new
> Jargon File.
To spread the y-series in the other direction, I have seen (used) `tayste'
(taste) for 2 bits. But of course, we all know two bits are a quarter.
--
| Frederick G. M. Roeber | e-mail: roe...@caltech.edu or roe...@vxcern.cern.ch |
| r-mail: CERN/SL-CO, 1211 Geneva 23, Switzerland | telephone: +41 22 767 5373 |
| "Why don't we nuke Iran, too, and say it was just a typographical error?" |
Sameer Parekh
I had thought the definition of a byte was the storage necesary to
store a letter of the alphabet. . .
(Therefore a Chinese byte would be some 10-11 bits :-)
Phil Kernick
The following definitions are a combination of real and suggested names
(the ones with an number next to them described below)
bit
(1) crumb 2 bits
nibble 4 bits
byte 8 bits
word 16 bits 2 bytes
longword 32 bits 4 bytes
(2) sentence 64 bits 8 bytes
(3) paragraph 128 bits 16 bytes
(4) page 2048 bits 256 bytes
Footnotes:
1. From Jargon file 2.1.5
2. I made it up, but what else is between a longword and a paragraph?
3. From the Intel segmented architecture.
4. This seems to have changed in the last 10 years. When I first
started playing around with 8bit microprocessors (6502, 6800)
this was definitely the definition, but now with 16/32 bit
processors it seems more common to have:
page 8192 bits 1024 bytes
but I feel that history should *not* be mucked about with.
Does anyone else have definitions that go between a paragraph and a page?
Maybe some of this could go into the updated jargon file...
Phil.
--
/// Phil Kernick EMail: ph...@adam.adelaide.edu.au
/// Departmental Engineer Phone: +618 228 5914
\\\/// Dept. of Psychology Fax: +618 224 0464
\/// University of Adelaide Mail: GPO Box 498 Adelaide SA 5001
Mark Zellers
>
>At least some of the Harris machines do (did?) this. One that my husband has
>had to transport data from I believe uses 6 bits/byte, and also uses weird
>type declarations in Fortran like REAL*6 and INTEGER*3. Truly bizarre.
I worked for Harris for two years after graduating from college. I was
in the OS Sustaining group, working on the H-x00 series. They were 24
bit (word addressable) machines. Bytes were 8 bits, it was just that
there were 3 of them to the word rather than the usual 4. This made for
an interesting trade off as far as integer performance was concerned.
While 16 bits is not large enough in many cases and programmers are
forced to use "double" integers of 32 bits, 24 bits seemed to be enough
that you didn't have to resort to double precision as often.
Of course, this plan did run into trouble when working with character
data rather than integer or floating point numbers. To construct a byte
pointer involved dividing by 3 rather than simpler shifts and rotates.
The other interesting thing about these machines was that the
instruction that moved values from one register to another was actually
an OR. The result was that you could OR the A,E,I,J,K and T registers
together and store the result in any or all of A,E,I,J,K and T in a single
instruction! Apparently this could all happen in a single cycle on some
of the early implementations, but was not such a big win on the
later microcoded processors. I forget the details, but "T" was a
special timer register. A was the main accumulator, E was the
"extension" register. I, J, and K were index registers.
Mark H. Zellers
decwrl.dec.com!voltaire!bwayne!mark
Gary Koerzendorfer
In one of my classes [customers, not necessarily college students], I was
informed that "tribble" was in use in the South [USA] for an octal digit,
a reference to the 3 bits quantity. Haven't heard it elsewhere, and this
was years ago. Anybody else heard it??
Start a different basenote/subject if you want to flame octal,
Gary
ga...@hpsid.sid.hp.com disclaimer: your mileage may be in kilometers.
Brian Cooper
>
>The following definitions are a combination of real and suggested names...
[summary: bit, crumb, nibble, byte, word, longword, sentence, paragraph, page
>Does anyone else have definitions that go between a paragraph and a page?
I remember a spoof proposing a computer architecture based on three-value
logic (trinary) rather than two-value (binary). The three logic states
would be yes, duhhh, and no. The smallest element of memory would be a
trit, three trits would be a trite, three trites were either a cliche' or
a trilobyte; I don't remember which. The article was fairly lengthy and
rather well-written; anyone else seen it?