I find myself largely agreeing with Carl on this:

> The use of a particular terminology is not the same kind of question as
> the flatness of the Earth.  One is truth, the other is just a matter of
> convention and agreement.

This is exactly correct. Anyone who knows SI knows it is hardly 100%
consistent and perfect -- it's just a convention that, if everyone uses,
will help reduce the number of units in use. Yes, it's a better convention
than most historical measurement systems, but it is just a convention.
> The fact is, the use of powers of two *is* the correct way to build
> memory structures in computers, if you accept that the simplest and
> cheapest way is the correct one. . .  But no computer
> engineer will misunderstand someone who speaks of a "32 kilobyte L1
> cache".  They will know that it is 32 * 2^10 bytes.

Carl is again correct -- no one who actually designs and builds products
using computer memory has any confusion when using "metric" prefixes for
binary numbers.
While I find the new "binary" prefixes interesting, I doubt they will ever
find any substantive usage. They remind me of the IEEE standard of many
years ago to replace the then-common symbols for logic functions (AND,
NAND, NOR, etc.) with rectantular boxes and a symbol for the function
(e.g., the ampersand for the AND function).
When it first came out, Texas Instruments started using it, but no one
else did. And, as far as I know, it is pretty much forgotten today. Those
"then-common" symbols are still the ones used today. And that is what I
expect will happen with these "binary" prefixes.
Jim Elwell


Reply via email to