On Mon, 21 Sep 2015 16:53:12 +0200 Alexandre Ratchov <a...@caoua.org> wrote:
> On Mon, Sep 21, 2015 at 02:29:03PM +0000, Christian Weisgerber wrote: > > On 2015-09-21, Stefan Sperling <s...@stsp.name> wrote: > > > > > The function that parses funny numbers is iswdigit() which gets a wchar_t. > > > But sleep(1) doesn't need that. > > > > The sole somewhat realistic use of i18n in sleep(1) is the decimal > > separator, so you could do > > > > $ sleep 1,5 > > > > in an appropriate locale. Of course the current code doesn't support > > that. > > IMO commands syntax should be considered as "computer language" and > shouldn't depend on the locale. > > In base commands, locale is only meaningful to generate strings > that won't be parsed by the machine, like text of error messages. > > Здравствуйте, Localised error messages in Russian: $ humppa ksh: humppa: не найдено $ кто humppa кто: humpa: Команда не найдена. No really, even native speakers do not want 'misfortunate' translations and the ambiguity this adds: what the hell were they thinking 'losting it into transition'... I don't want error messages or any feedback produced by the machine and the operating system in any other language than English in C locale, even if my locale is set to native (non-English). So, probably you can save the trouble wrapping around such functionality and consider test of error messages also "computer language" suitable for (easy) machine parsing. Another one of those popcorn threads (sorry I'll not post again here just could not resist to add a non native speaker's opinion about butchered 'transversion'). С уважением, Антон