i18n question.

[Kenichi Okuyama]

Dear Alex,

>>>>> "Alex" == Alexander Bokovoy <a.bokovoy at sam-solutions.net> writes:
Alex> Basically, if you have system with inconsistent charset usage between
Alex> filenames and user/group names, you'd already screwed. :) On other hand, 
Alex> I see no problem in having user/group names in non-ASCII but in the same encoding 
Alex> used to encode all other components (file names, etc).

If you are building system from scratch, you are right.
Problem occurs when you have Legacy system, and have to port those
user to new system.
# Yes 'Legacy' is where trouble begin. Always...


Assume you were using unix on EBCDIC machine (don't ask me why).
Your username and password was managed in EBCDIC. You have 10k users
on your machine, and you know that password entry encoding uses the
EBCDIC character code for generating hash.

One day comes, when you have to move your men to Linux. You find
filesystem can be moved easily to UTF-8, but password .... You have
to decide ether to add 'EBCDIC password' module, or ask 10k men to
restart their password, or crack 10k passwords and convert them to
ASCII password.
# Why not simply convert EBCDIC password to hash value then
# re-calculate ASCII string for it? Well, that's because "Hash
#value" is being effected by EBCDIC char code. Same password in
# ASCII and EBCDIC usually ends in different "Hash value".

Do you blame yourself if you select 'EBCDIC password' module
strategy?


Tragedy, though......

best regards,
---- 
Kenichi Okuyama