Re: CVS commit: src/bin/sleep

[christos%astron.com@localhost (Christos Zoulas)]

In article <9790a289-8b3e-e57a-e4d8-c021aa7b1b38%gmx.com@localhost>,
Kamil Rytarowski  <n54%gmx.com@localhost> wrote:
>> 
>> And as I said before, if this is to be changed, it needs to be via
>> a discussion in front of a wider audience than reads source-changes-d
>> and in particular messages with a subject that refers to one of the more
>> boring commands that we have.
>> 
>> kre
>> 
>> ps: in the meantime, I have changes to sleep which avoid the horrible
>> '.' test, that I was never happy with, don't alter sleep's locale (using
>> strtod_l as Joerg suggested --- why is that not documented?) and
>> deal mor sensibly with long sleeps (nanosleep(2) is just plain weird when
>> asked to sleep for a long time - sometimes it returns an error, other
>> times it simply returns(0) immediately ... try sleep 1e15 for example).
>> Those are not yet committed.   But probably will be.   Most of it has
>> nothing to do with any locale issues.
>> 
>
>I have really met a person (after computer studies) who found a command
>line tool broken as the comma separator was '.' instead of a local ','.
>Handling both is acceptable... however it has serious issues in other
>tools that accept comma like sort(1). Supporting locale specific
>separator will create trouble in making scripts portable.
>
>I think it's better to keep an uniform way of handling separator in
>basic command line tools and restrict locales to translations only.
>
>My native locale uses ',' for separator... but everybody is acquainted
>with . as it's everywhere (in calculators etc). I think this is a losing
>battle to keep ','. I personally always used dot in schools etc and
>nobody ever complained.

I think that we should consult what the standards say about all of this,
what is the current practice on other OS's, and then decide and document
the decision and approach to handle this consistently across all applications.

christos