Date   : Tue, 28 Jul 2009 00:05:21 +0100
From   : pete@... (Pete Turnbull)
Subject: Units of memory, Was: Master Ethernet upgrade

On 27/07/2009 22:38, Jonathan Graham Harston wrote:
>> Message-ID: <4A6C801B.4020301@...>
>  
> Pete Turnbull wrote:
>>>> Jonathan Graham Harston wrote:
>>>>>  1K is 1024 bytes. > >>
>>>>    ^--------+ > >> I would imag|ine
>>>> thus 1KB = 1024 bytes? > >
> |  ^-----------------------+ No, 'cos that-+ is using 'K' as
>>> a unit, but that-+ is using 'K' as a multiplier. 1KB is
>>> 1*1000*B and B is 1 byte. But K isn't a unit, even in
>>> computer science, unless of course you're
>  
> Technical terminology embedded in computing science for over 60
> years. Are you going to go back in time and reprogram the BBC to
> ensure it starts up saying "32.768KB" ?

I didn't write 1000; Rick did.  You've trashed the line wrapping.  I 
meant 1024, not 1000, and wrote:
 > one kilobyte is 1KB (and 1Kb is 2^10 bits).  Your usage is slang,
 > and not correct.

> One "1K" as a unit of memory size - as opposed to "1KB" - always
> was, always is, and always will be 2 ^ 10, and 2 ^ 10 always was,
> always is, and always will be 1024.

I agree :-)  Kibi-lovers notwithstanding.

> Pete Turnbull wrote:
>> On 25/07/2009 16:40, Jonathan Graham Harston wrote:
>>> No, 'cos you can only have one multiplier prefix. It should have
>>> had a ;) as memory units never take prefixes, you use a different
>>> unit.
>> That's not true either.  Using several multipliers is quite common in
>> electronics and physics, and perfectly acceptable in the SI system.
>  
> Are you sure? Why would you use mmA when there's a perfectly usable
> uA?

You wouldn't, but milli-micro-farad is still accepted in some circles, 
even if seriously archaic.  It's long been accepted that, especially for 
smaller multipliers, if the "proper" multiplier is either not well 
known, or not invented, you have to use two or more together.  That's 
not in strict compliance with the SI rules (I was wrong about that, I 
checked BIPM), but that's not the only game in town, and our use of Kilo 
(K), mega (M), and giga (G) breaks the rules too (but interestingly, the 
official rules don't actually mandate the Ki/Mi/whatever prefixes instead).

More to the point, memory units do take prefixes.  Always have, in my 
world (and IBM's, which is where the currently-accepted definition of a 
byte came from).

> Digging out my Tennent's, p37 says: "To obtain multiples and
> submultiples of units, //a// standard prefix is used, as shown
> below" (my emphasis)

Sure.  Wherever practical or sensible.  Not, however, an absolute 
requirement in all circumstances, where the appropriate prefix doesn't 
exist (not very common these days, but possible).  The last few prefixes 
were only added to the official list in the mid 1990s.

-- 
Pete                                           Peter Turnbull
                                               Network Manager
                                               University of York