Re: Bad Metric ... 1k = 1024 or 1000?

[Rubywand <rubywand@swbell.net>]

David Empson writes ...
> 
> Rubywand <rubywand@swbell.net> wrote:
>
> > Agree. 1k bytes, Ohms, Volts, eggs, ... is, precisely speaking, 1000 of
> > them. That we expect "1k bytes" to be 1024 bytes is beside the point. The
> > accepted interpretation of "1k bytes" is that it is a term of convenience
> > which incorporates a roundoff.
> >
> > Similarly, the normal interpretation of 140k as the capacity figure for
> > a 5.25" 16-sector Apple II diskette side is that it is a nearest 10,000
> > roundoff designator. A similar sort of understanding exists for "4MB
> > RAM".
> 
> But it isn't always a "nearest 10000" roundoff.  An "800K" disk holds
> 819200 bytes. If the disk capacity was rounded to the nearest 10K, then
> we would call it an 820K disk. 
 ....

     Yes. That is why my statement says "A similar sort of understanding exists
for '4MB RAM'"  rather than  "The same applies to ... ", etc.. Since, for A2
diskettes, "140k" is generally understood to be an approximation for 143,360,
the inferred roundoff is likely to be 10,000. The magnitude of the inferred
roundoff depends upon the magnitude of the quantity being rounded.


> 
> Almost everyone in the computer industry uses the definitions of K =
> 2^10, M = 2^20, G = 2^30 and T = 2^40, except hard drive manufacturers,
> who use the normal metric definitions.

     No problem. You are talking about naming conventions or 'terms of
convenience'. To the extent that this usage conflicts with established standard
units, there is, naturally, a chance for confusion.


 
> For example, RAM is always described in manufacturers' data sheets using
> the binary K/M, unless they describe it as an exact number of bits or
> bytes (e.g. "131072 bits"), which gets unwieldy when you are talking
> about something like a 64M RAM.

     Some data sheets are better written than others. "128k x 1" is a better
description than "131072 bits" because '128k x whatever' is the terminology
most designers are used to.



> > Notwithstanding the above, those who argue that, for memory, disk, file,
> > ... size, "1k" is a standard unit = 1024 can find some support in the
> > way file size is reported by popular operating systems. For example, GS
> > System and Windows 95 both use 1k = 1024.
> 
> So does the Mac.  Not sure about Unix (I don't recall seeing any
> abbreviations - it displays file sizes in directory listsing as an exact
> number of bytes).
>
> Does anyone know of any operating system on any platform which uses K, M
> or G abbreviations, and which defines them as being 1000, 1000000 and
> 1000000000 instead of 1024, 1048576 and 1073741824?
> 

      Do not know of any-- seldom use anything but Apple II OS's and Win95.
That, in 1999, OS's still commonly use 1024 for "k" just illustrates an
instance of sloppiness-- a flaw needing to be corrected.

      A file catalog is not like a data sheet RAM size description. It is
supposed to tell the user how big each file is. Hardly anyone routinely thinks
in hexadecimal units using decimal numbers or is good at multiplication or
division by 1024.

      Whenever system information is reported to a user, it should be in the
most easily understood format. In some cases-- e.g. memory addresses-- this may
be in hexadecimal. Obviously, something as mundane as file size should be
expressed using everyday, easily interpreted units.




Rubywand