Re: address@hidden: Re: [PATCH] I18n flag for msgfmt]

[Paul Eggert]

Behdad Esfahbod <address@hidden> writes:

> * Diff is a text utility, and <b>when using diff with Persian
> messages</b> better you get Persian digits.

That's a reasonable request, but it conflicts with a POSIX requirement
and with longstanding tradition and generates patches that "patch"
doesn't grok, so my kneejerk reaction is that it should be enabled
only by a new option.

> You already get the date in Persian.

This is no longer true in recent versions of GNU diff, partly because
of the compatibility hassles mentioned with 'patch', partly because of
the portability hassles of generating a proper localized date on all
platforms.  Instead, you get the date, time, and time zone in ISO 8601
format, e.g. "2003-09-01 11:44:52.771112000 -0700".  This occurs
whenever you specify any locale other than "C" or "POSIX", as POSIX
requires that these use the traditional English format.

With other versions of diff it does make sense to use LC_ALL=C to make
patches, but this matters only for recipients who need to parse the
time stamps (i.e., "patch -Z" or "patch -T").

> I can't find any reasonable example that you want to write such a
> weird variable type (off_t, ...) with Persian numerals.

This is clearly a matter for Persian translators to decide (not me!).
But to see some examples where it might come up, you can look for
instances of "offtostr", "human_readable", and "PRIdMAX" in standard
utilities.  Here's one example: when dd reports "advancing past
21343244 bytes in output file FOO" after an output error, the number
21343244 is an off_t value.