Re: [bug-gnu-libiconv] Problem found in MSDOSFS of FreeBSD

[Bruno Haible]

Hello,

Thor Ablestar wrote:
> There is a problem. Russian MS Office quite often generates files whose 
> filenames contain character "No." № (UTF code 0x2116). ...
> the  FreeBSD team informed me that for now the library that converts filename
> encodings is FreeBSD port of libiconv ...
> It appeared that UTF to KOI-8R conversion tables do NOT contain code 
> 0x2116.

Yes.

According to Wikipedia [1] the VFAT file system can contain file names
with arbitrary Unicode characters.

The character U+2116 is available in some legacy encodings:

DOS     CP855       0xEF
IANA    CP866       0xFC
DOS     CP1125      0xFC
Windows CP1251      0xB9
ISO     ISO-8859-5  0xF0
MacCyrillic         0xDC
MacUkraine          0xDC

> Also, the CP-1251 locale works but I will shoot myself twice if somebody 
> forces me to use it. The same with CP-866 locale. Both have this code on 
> proper place, and both have "No." symbol in place.

Yes, using CP1251 or CP866 as a locale encoding in 2012 would be
extremely backwards-directed.

Modern Unix distributions (GNU/Linux, MacOS X, Cygwin for example) are using
UTF-8 as locale encoding for most locales. It would make sense to improve
the UTF-8 locale support in FreeBSD until you find it usable enough.
(NetBSD also has UTF-8 locales, but I don't know how usable they are.)

Bruno

[1] http://en.wikipedia.org/wiki/VFAT#VFAT