On Wednesday 13 June 2007 15:29, Scott James Remnant wrote:
> On Wed, 2007-06-13 at 12:51 +0200, Christof Krüger wrote:
> > On Tue, 2007-06-12 at 15:52 +0100, Ian Jackson wrote:
> > [...] Please tell me the disadvantages so there can actually be a
> > constructive discussion.
>
> User Confusion.
>
> Most users do not know what a "tebibyte" is, and they do not care. They
> know that "a terabyte" is "about a million million bytes", and that is
> sufficient.
>
> Since you're rounding anyway, the loss of accuracy between "about a
> million million bytes" and "just over a million million bytes" is not
> significant. Certainly not at the expense at having to teach users
> another new unit.
This is a hurdle to adoption, a one time cost. It is not an argument against
IEC prefixes per se. The long-term benefits outweigh the costs, IMO.
> Hard drives are bought in gigabytes, memory is bought in gigabytes, etc.
> Quoting the same figures with a different unit in the operating system
> is pedantry for its own sake.
>
> Users have already learnt that the term "gigabyte" is approximate.
No, it's not. It's ambiguous. A given number can be exact or approximate
regardless of the unit. "1.0 GB" can mean either 1.0·10^9 byte or 1.0·2^30
byte, but whether the real value is exactly one or the other or something
near one or the
> Introducing new units has only added confusion, rather than removed it.
New concepts can always cause initial confusion. Relearning is harder than
learning something right from the beginning. The same argument has been used
against metrication in the US. Again, it's a one-time cost.
> Before the new units, we all knew that 1GB was an approximate figure and
> likely to be (for bytes) based on a power of 2. Now we have figures
> quoted in GB and GiB, some of which are power of 10, some of which are
> power of 2. Some figures quoted in GiB are wrong, and should be in GB;
> likewise some in GB should be GiB. And we still have many figures in
> both GB and GiB which are neither of the two!
You're talking a lot about approximation. If I understand you correctly,
you're saying that any stated quantity of data must either be an exact
number, e.g. 23 368 986 120 bytes, or an approximation with a single
significant digit. That is *stupid*. You want to deny people the
*possibility* of consistence, unambiguity and accuracy (without resorting to
numbers on the form "3.1·10¹²"), just because you won't think that you'll
need that possibility most of the time.
There *is* reason to state rounded numbers with two or three digits, in which
case the difference between MB and MiB or GB and GiB definitely matters, and
even with a single significant digit, 8 GiB (exactly) is 9 GB when rounded to
the nearest whole number.
> Renaming the 1.44MB floppy helps in neither case; it is neither 1.44MB
> or 1.44MiB. One could name it the 1.4MB or 1.47MiB floppy and confuse
> everyone into thinking it's a different thing, of course. Or maybe it
> should be the 1,440KB floppy, or the 1,475KiB floppy? Neither of these
> help the situation.
The 1 440 KiB floppy is dead. Let it rest in pieces. The fact that a marketing
department screwed up long ago by thinking that 1 440 kB equals 1.44 MB,
which it would have done, had that really *been* 1 440 kB and not 1 440 KiB,
is not a case against IEC prefixes. On the contrary, it may well be a prime
example of a confusion that wouldn't have happened if the IEC prefixes had
been adopted by then.
--
Magnus Holmgren holmgren@lysator.liu.se
(No Cc of list mail needed, thanks)
Attachment:
pgpUXEhYlqj8k.pgp
Description: PGP signature