Re: [PATCH 1/2] posix: User scratch_buffer on fnmatch

[Florian Weimer <fweimer@redhat.com>]

* Bruno Haible:

> Paul Eggert asked:
>> > By the way, how important is it to support awful encodings like
>> > shift-JIS that contain bytes that look like '\'? If we don't have to 
>> > support these encodings any more, things get a bit easier.
>
> Here we are talking about locale encodings, and Shift_JIS (as well as
> SHIFT_JISX0213) are not usable as a locale encoding in glibc. See e.g.
> [1], [2].
>
> That's the reason why no Shift_JIS locale is listed in
> glibc/localedata/SUPPORTED. [3]

> [1] https://sourceware.org/bugzilla/show_bug.cgi?id=3140
> [2] https://sourceware.org/legacy-ml/libc-alpha/2000-10/msg00311.html
> [3] https://sourceware.org/git/?p=glibc.git;a=blob;f=localedata/SUPPORTED

We used to have a fully supported product based on the original
Shift-JIS.  It did not require glibc changes (we package both localedef
and the locale sources, so it's easy to build custom locales), but other
GNU components had to be patched.

> Florian Weimer wrote:
>> There is a Shift-JIS variant which is ASCII-transparent (Windows-31J,
>> it's also specified by WhatWG/HTML5), so from a glibc point of view, it
>> would be just an ordinary charset like any other.
>> 
>> But feedback we have received is that the users who want Shift-JIS
>> really want the original thing.
>> 
>> We do not presently support either variant downstream, but one potential
>> way forward would be to turn Windows-31J into a fully supported glibc
>> charset with a corresponding ja_JP locale (which would imply downstream
>> support as well), and just hope that it displaces the original Shift-JIS
>> in the future.
>
> I don't think there's a real need for that. In the years 1995 ... 2005
> there was a lot of resistence against Unicode in Japan, because
> Unicode maps several slightly differently looking glyph images to the
> same glyph/character (even for Western encodings, for example the
> Polish accents look a bit different than the French ones), and - at
> the time - Unicode did not have means to disambiguate these, thus
> people complained about "characters are rendered incorrectly if you
> use Unicode". This has been resolved for more than 10 years already.

We saw commercial demand for Shift-JIS much later than that.  I think an
official Windows-31J-based ja_JP would still be welcomed at this point.

A Windows-31J locale could be added to localedata/SUPPORTED.  We have
not done that yet because someone wanted to look into alignment between
Windows, HTML/WhatWG and what we currently have in the source tree, but
that hasn't happened yet, unfortunately.

Thanks,
Florian
-- 
Red Hat GmbH, https://de.redhat.com/ , Registered seat: Grasbrunn,
Commercial register: Amtsgericht Muenchen, HRB 153243,
Managing Directors: Charles Cachera, Brian Klemm, Laurie Krebs, Michael O'Neill