Re: [Gnu-arch-users] thoughts on utf-8 filenames

[Tom Lord]

    > From: Johannes Berg <address@hidden>

    > Actually, I think it boils down to:
    > Does tla want to be `smart' about filenames, ie. convert them, or does
    > it just want to store the byte-sequence in filenames as-is?

I think that the longer-term answer is that it converts them, using a
(perhaps extended) UTF-8 format internally, in changesets, in tar
files, etc.

The ideal is, for example, that you can locally be using ISO-8859-8
filenames and I can locally be using UTF-8 (but limitign myself to
characters that also occur in ISO-8859-8) -- and we can interoperate
just fine.   Well... I guess we just hope that somewhere in the
toolchain our "#include" directives and the like are compatibly
transcoded :-)

To keep the door open to that longer-term answer, for now it's best to
limit filenames to 7-bits characters -- and if someone has a
short-term need for a non-public archive using 8-bit characters, they
can make the trivial patch to enable them.

-t