On 10/22/2014 10:00 AM, John Eells wrote:
> [email protected] (Chase, John) wrote:
> <snip>
>>> * I believe that is most likely GiB for the pedantic among us.
>>
>> IIRC, DASD capacities are expressed in decimal, rather than binary,
>> quantities:
>>
>> 4KB = 4,000 bytes;
>> 4MB = 4,000,000 bytes;
>> 4GB = 4,000,000,000 bytes;
>> Etc.
>
> This is perhaps the strangest marketing decision ever made by the
> industry. I don't even know who started it (IBM or a PCM way back
> when--it certainly predates me). But while it's true that we advertise
> (unformatted, IIRC) disk space in powers of ten, most system limits
> are in powers of two, including this one.
>
Actually, I always thought it made perfect sense for DASD capacities to
use decimal conventions.  The capacity in characters/bytes per track,
number of tracks per cylinder, and number of cylinders per device for
early DASD were based on physical attributes of the device having
nothing to do with powers of 2.   Decimal notation and power-of-ten
multipliers were a natural choice.    Computer architectures existed in
1950s and 1960s that didn't use binary addressing of main memory either,
and those used decimal suffix conventions for describing memory capacity
as well.  It was only on machines where the hardware instruction
architecture used binary addressing and memory increments were a power
of two that KB, MB, and later GB became "corrupted" when talking about
memory space to mean what should now be called KiB, MiB, and GiB. 

Although today's hard drives may typically have a fixed-block
architecture with a block size that is a power of two,  empirical
evidence shows it is still the case that the number of blocks per track
and cylinders per device is still constrained by physical properties of
the drive that have no relation to powers of 2, so it still makes sense
to continue to describe capacity with decimal multiplier suffixes. 

What I instead still find extremely strange is the decision with
MVS-z/OS to go with binary power multipliers for space values in SMS,
JCL space allocation by records, and ISMF space displays, which went
counter to long-established DASD space conventions.   The size required
for a typical data set is mainly a function of logical record size and
maximum number of records, neither one of which quantity is constrained
to a power of two and most certainly would be described by the end-user
community using decimal values, not values with binary multipliers.

At the vary least, since the introduction of standard binary suffixes in
1998, one should consistently use the correct suffix notation whenever
possible, and especially to avoid confusion where binary memory notation
and non-binary DASD notations collide.  The VSAM file limit is based on
a binary addressing constraint on the total number of bytes in all CIs 
within the file and should be expressed as 4 GiB.

I would hope there is at least a goal in place to gradually make all IBM
documentation and software displays compliant with the 1998 suffix
standards to avoid ambiguity and confusion to newcomers unfamiliar with
the confusing contextual interpretation of these suffixes that was
required before 1998, but which should now be avoided.

-- 
Joel C. Ewing,    Bentonville, AR       [email protected] 

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to