Re: [Gnu-arch-users] Increasing the filename space (Or: begging for trouble?)

[Tom Lord]

    > From: Aaron Bentley <address@hidden>

    > Ugh.  TLA has logs as files, not data.  So if the log file is written in
    > ISO-8859-2, TLA should not rewrite the file as UTF-8.  

    > Perhaps make-log should supply a Content-encoding header based on the
    > local character encoding?  

Perhaps.

I think the current situation is close to right:  log files must use
an octet-based encoding with ASCII as a subset.   Headers added by tla
itself are purely ASCII.

A content-encoding header would be handy for output commands (e.g.,
`cat-log') and in browsers.  Given a log file with an encoding header,
and a desired output encoding, it's (mostly) a simple matter to read
the log, _perhaps_ unescape the tla-generated headers, and transcode
to the desired output format.

Well, _mostly_ a simple matter.   We don't actually know the encoding
(after unescaping) of filenames, and that's a nastier issue.   Not
just for catting logs but in general.

As an example, I might have 8859-4 filenames on my filesystem and you,
utf-8 (but, since we work on the same project, using only 8859-4
compatible characters).   In this case -- you and I have a mess.

Hmm.... hold on: lemme put the rest in a separate message....


-t