On 2015-09-21 16:45, Mark Kettenis wrote: >> From: Christian Weisgerber <[email protected]> >> Date: Mon, 21 Sep 2015 14:29:03 +0000 (UTC) >> >> On 2015-09-21, Stefan Sperling <[email protected]> wrote: >> > You could argue that the thousands separator should be supported though: > > $ sleep 1.000.000 > > if your locale is something vaguely european, and > > # sleep 1,000,000 > > for the north-americans. > > But let's not go there...
In my locale (Sweden), the thousands separator is space... try to meaningfully parse sleep 1 000 000 without all of a sudden introducing mandatory quoting. :-) So no, let's not go there... Like many others I've seen plenty of trainwrecks caused by mindless internationalization in other operating systems, and how it wreaks havoc in scripts and interpretation of piped data. It is NOT pretty. The stance that the base system should only speak English and use time-tried and known formats is IMO the correct one. With that said, we do live in an international world and I was alarmed about something Stefan wrote earlier in this thread: > When we still had latin1 etc. it was possible in theory that values between > 128 and 255 represented a digit. But now, isdigit() does the same regardless > of locale setting (C or UTF-8) since it cannot be given a multibyte sequence, > i.e. it will not deal with character values above 127. Maybe I'm the only one that this is news to (and I admit I obviosly haven't followed tech@ and misc@ closely enough lately), but am I interpreting this correctly in understanding that 8-bit locales except for the "C" locale (which is effectively a 7-bit locale) are gone? That would be extremely unfortunate, since UTF-8 (even if it was fully implemented in OpenBSD which of course it isn't) is *not* the sole answer to the world's localization problems. At least not in the forseeable future. For example, I still have plenty of systems running that absolutely relies on a working ISO 8859-1 environment, and not just in the sense that it is 8-bit transparent but in understanding collation orders, proper case folding and functioning of the is...() macros/functions etc. I tried to google for more on this issue, but came up pretty thin. Has this been discussed publicly somewhere? If this is due to shortage of manpower or interest or something like that, maybe we can do something about that. I'm sure it isn't still a matter of lack of knowledge of the problems facing us non-native English speakers - since I know OpenBSD developers come from all corners of the world - so please someone tell me I've misunderstood, or that it's just a bad dream and I'll wake up soon. :-) * * * Are there more people like me who really could use more complete i18n support in OpenBSD (multibyte as well as legacy 8-bit) and are willing to make efforts into implementing it? Is there a coordinated effort somewhere? If not, let's discuss it! Offlist is fine of course if it isn't of general interest (but I think it is). And by "implementing", of course I mean doing it right, the OpenBSD way. If there is interest but no coordination, I can start it off and try to make an inventory of the current state of implementation and what efforts may already be ongoing (or dormant or abandoned, due to lack of interest/time/manpower/resources/pizza/beer/hikes). What say ye? * * * Regards, /Benny
