>1) A multiuser machine, with users using different charsets. > Who decides which one is "local"? > >2) The sysamin/user changes the charset, e.g. from iso-8859-1 > to iso-8859-15 to get the Euro character. > How should the filenames stay in the local charset when > this changes? Would there be some automatical conversion?
There are problems, yes. What you have failed to show is that your solution is better, or even implementable. >A non-broken solution will have to convert charsets somewhere >between the filesystem level and output to the user's terminal. >(And no, I don't know an easy way to do this :-( ) Converting a byte-string as if it were a string of characters is guarenteed to cause problems. There will be unaccessable files, multiple files with the same name, all sorts of problems and security holes. Not to mention you have to rewrite every piece of code that handles filenames. Good luck. The non-broken solution which everyone else is going towards is complete conversion of the system to UTF-8; most programs already support UTF-8, and once the switch is done, it will be clean, without the breaking of POSIX rules or adding more code to every program.
