Re: [gentoo-project] Call for agenda items - Council meeting 2014-04-08

[Joshua Kinard]

On 03/30/2014 11:47, Micha³ Górny wrote:
> Dnia 2014-03-30, o godz. 09:56:45
> Joshua Kinard <kumba@g.o> napisa³(a):
>
>> On 03/30/2014 05:23, Rich Freeman wrote:
>>> On Sun, Mar 30, 2014 at 4:33 AM, Micha³ Górny <mgorny@g.o> wrote:
>>>>
>>>> ISO/IEC prefixes [1,2]: KiB (kibibyte), MiB (mebi-), GiB (gibi-)
>>>>   -- unambiguously 2^10, 2^20, 2^30
>>>>
>>>> 'old' prefixes: kB (kilobyte), MB (mega-), GB (giga-)
>>>>   -- can mean 10^3 or 2^10 etc. depending on author's intention
>>>>   -- SI people tend to use 10^N for consistency with other units
>>>
>>> To think that counting is the one situation where it is actually
>>> possible to have perfect precision, and we still manage to mess up the
>>> units...
>>>
>>> I'm a bit torn on this.  I'm an American, so I am supposed to do
>>> something ambiguous and arbitrary (can we count bytes in baker's
>>> dozens?).  I'm also a Chemist, and never met an SI unit I didn't like.
>>>
>>> We do need to keep in mind that the laws of physics dictate that
>>> boolean states can only be combined in groups of 8.  That's a shame,
>>> because kibibits just rolls off the tongue and helps promote healthy
>>> fur.
>>>
>>> Now I need to get back to sleep because I have a 97.7 kibimeter drive
>>> tomorrow...
>>
>> I'm on the side of using the old prefixes.  I just never groked the base-10
>> forms.  Since Gentoo is all about choice, why not make it configurable?
>> Maybe a USE flag or some other switch in make.conf that can be read by
>> Portage to let you choose if you want natural computer size presentations
>> (old prefix) or the human-friendly ones (new prefixes).  We've obviously got
>> the code for both forms in the referenced eclass, so just wrap it in a
>> conditional based on the chosen switch.  That laves the debate up to which
>> value is the default.
>
> Don't you think this would be a serious overkill?
>
> If choice for this were to be provided, it should probably land
> in glibc's LC_MEASUREMENT. Though I'm not sure if that would help
> something since most of the countries seem to support the binary
> prefixes. At least the countries using the metric system.

I'll take overkill versus endless bickering over what is really an extremely
minor issue.  If there's already some kind of variable provided by the
system libc (be it glibc, eglibc, uclibc, or freebsd-libc) that can decide
on the preferred measurement system to use, then yeah, we should use that
instead of building our own version.

The "how" we do it is more of a technical thing, and not really what is
being discussed in this thread.  If I knew more about LC_MEASUREMENT and the
code in the affected eclass, I'd probably try to hack up something myself
and propose a patch to -dev, but I think others here, who have more
knowledge on the topic, could work up something much better.

The "what" we do is the point of contention in this thread, and I think
rather than starting a holy war over measurement systems (as sometimes
happens between the majority of the world that uses metric, and the three
countries that use imperial, including the US), just allow for the choice of
both forms, and decide on a sane default.

Personally, I say make the metric version the default (new prefixes), and
leave the choice available to switch to the SI (old prefix) form with a
simple variable switch in a config file some where.

Now, I'm off to prove that black is white, then investigate this whole
"zebra crossing" thing...

-- 
Joshua Kinard
Gentoo/MIPS
kumba@g.o
4096R/D25D95E3 2011-03-28

"The past tempts us, the present confuses us, the future frightens us.  And
our lives slip away, moment by moment, lost in that vast, terrible in-between."

--Emperor Turhan, Centauri Republic