Confusion over encodings, utf-8, etc.

[Bill Hacker]

Chris Csanady wrote:

On 8/29/06, Jonathon McKitrick <jcm at xxxxxxxxxxxxxxxxx> wrote:

I've entered the world of unicode with my mac, and I'd like to make 
things
consistent across my machines.


Are you interested in utf-8 for filesystem encoding as well?  This is
where utf-8
becomes painful.
Yes - it almost justifes Linux.  But not quite. D'ruther have file loss...

 It turns out that there are several normalization 
forms for
it, where Windows uses NFC, and and MacOS uses NFD.  The difference
being that identical characters can be stored in different ways.  For 
example,
for a character with an umlaut, the umlaut can optionally be external.  See
http://www.unicode.org/reports/tr15/index.html for more information.

ACK.

Now, the mac expects NFD, and the Finder is completely incapable of
dealing with NFC.
Doesn't seem to be a 'deadly' problem on UFS mounts. Finder handles, for example 
Chinese filenames in UTF-8, displays correct Chinese characters.
(I haven't had an HFS or HFS+ mount for years).

The 'terminal' CLI shows them, OTOH, as escaped numeral sequences, as does my 
editor-of-choice (joe).

My current workaround is to use 'TextEdit' - set to UTF-8, to work with such 
files, ELSE a WP (SO/OO, NeoOffice).

> Trying to move those files around over NFS can result
in them simply dissappearing!  Likewise, NFD encoded files visible over
SMB on a Windows machine are not displayed properly.
Never saw the usefullness of NFS.  The only fs I use besides UFS is 
HPFS/HPFS-386, so SMB is the better tool for me, as the OS/2 and MSDOS NFS is 
even more broken than normal...

Unfortunately, the only portable option is to use SMB exclusively to export
NFC encoded files from a unix box.  In this case, the SMB client on the mac
handles the renormalization.  If you use NFS exclusively with a mac, note
that some non-native applications like Azureus will still write NFC encoded
filenames which must me converted manually.
ACK, but not an issue here.

If you have lots of files to translate, there is a convenient tool 
available in
pkgsrc: converters/convmv.  It allows you to recursively renormalize or
transcode a set of files from/to anything supported by iconv.  It needs 
to be
used over a fs which does not mangle characters though, so something
like NFS or a local filesystem.

Chris
Well -we all hope the DFLY project will help make soem of these issues less 
painful.  On OS X, perhaps ZFS, TVFS, or AFS could provide some relief? We have 
begin revisiting certain of the 'Plan 9' features, as it runs well as a guest on 
OS X.

Meanwhile we 'share' files when we must via restricted https, rsync, scp, etc.

'Sharity Lite' has helped.  Seems to be SMBFS at one end, NFS at the other, 
though it did mangle timestamps when backing up FAT and NTFS Winboxen to a UFS 
mount. Mounting the storage partition as MSDOS might have been all we needed to 
do to get aroudn that, thugh hardly a universal solution.

With rsync and scp, binary is binary, so the problem has to do only with 
filenames for us. Hong Kong, our HQ, is of course, very much prone to Chinese 
encoding - unfortunately seldom in UTF-8 anyway.

YMMV,

Bill