On Mon, Jan 28, 2019 at 07:45:12PM +0700, Robert Elz wrote: > | at least the Arabic ones (momayyez). Supporting 2-3+ styles is opening > | Pandora's box; > > No-one is planning that. There are two options - the C locale, or > the locale set in the environment.
Actually it's kind of what I was suggesting -- that shell programs that take floats should accept any string with an unambiguous parse in any representation style, and reject ones that do not. Although in some ways this: > (unless we someday make a command that works for LC_NUMERIC > the way iconv works for LC_CTYPE). is a better plan. Either way you want to be able to combine inputs that have different representations of the same data type, because in the real world people have colleagues from all over (just note the Cc: list on this email); they will from time to time send to you material using their local conventions, and you want to be able to process it without making a fuss or cursing them afterwards. The assumption of locales is that every user's data world is geographically small, and that wasn't true even when they were invented. -- David A. Holland dholl...@netbsd.org
