spc3r21c.pdf tupo kib != KiB

Pat LaVarre plavarre at lexarmedia.com
Thu Jan 20 13:39:41 PST 2005

* From the T10 Reflector (t10 at t10.org), posted by:
* "Pat LaVarre" <plavarre at lexarmedia.com>
"Standard" usage is split.

The device folk who actually pay to create the storage build their
companies on the physics of recording, be that flash, battery backup,
rotating magnetic, whatever.

Those physicists, like http://physics.nist.gov/cuu/Units/binary.html
now, have consistently defined T G M K defined to mean 10 raised to the
powers 12 9 6 3 since the metric era dawned, circa Bastille day 1789,
drawing its prefixes from Greek teras, Latin gigas, Greek mega, Greek
khiloi, etc.

Yes the vast majority of people involved are the customers who buy the
storage, and many of whom still speak of inches and ounces.

Yes the physicists would be unemployed, without the cooperation of those

Yes those customers have consistently defined T G M K in computing to
mean 2 raised to the powers 40 30 20 10, denoted instead as Ti Gi Mi Ki
by the physicists of the U.S. NIST but then also the international IEC
in December 1998, together with the pun of "mebibytes" pronounced like
"may be bytes".

But the physicists since the beginning have noticed that yielding in
this dispute would mean they lose 2.4%, compounded with every
generation.  To let T mean Ti costs 1.024 to the fourth power, i.e.,
more than 9.95%, i.e., fuggeddaboutit.

Our RAM colleagues could take a share of the blame here: in the
beginning, they sold RAM architected to end on boundaries that grew
exponentially as a power of two, without leaving room for ECC and other
overhead, as if physical storage were reliable.

I remember in 2004 some friends & I checked a small indeterminate sample
of flash - I can report that least those flash folk had begun trending
towards the physicists' T G M K, same as HDD always have.

* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org

More information about the T10 mailing list