Re: [Bug-tar] mac os x or general charset problem patch

[Paul Eggert]

Joerg Schilling <address@hidden> writes:

> POSIX describes a solution for your problem in case you are using a
> POSIX.1.2001 compliant tar archive.

It's not clear to me that his problem would be solved in general, if
he used such an archive.  That archive provides for representation of
names in UTF-8.  But it's quite possible that someone will want to
extract a UTF-8 name onto a file system interface that supports only
(say) Shift-JIS names.  That problem will resemble his situation: the
archive contains file names that cannot be stored in the file system.

So it seems to me that this is a general problem, one not specific to
Mac OS.  Admittedly it's a problem mostly with deficient file system
implementations like the one he's using.

A related issue here is that the POSIX 2001 convention for file names
is seriously broken on ordinary Unix-like platforms, in that UTF-8
cannot represent arbitrary Unix file names that contain bytes that
would be encoding errors in UTF-8.  (This is the inverse of his
problem: it's a file system that can represent names that POSIX 2001
UTF-8 cannot.)  If memory serves, GNU tar will sidestep this problem
by not using the POSIX 2001 UTF-8 names unless the user specifically
requests it.