Bug#99933: second attempt at more comprehensive unicode policy
On Thu, 2003-01-09 at 20:57, David Starner wrote:
> A Posix filename is a null terminated byte string (sans '/'). Any
> widescale conversion is going to cause aliasing issues and other
> bugs, whether or not we stay Posix compatible.
> Just as important, conversion is not an issue for debian-policy;
> linux-utf8@nl.linux.org (the primary Unicode-Linux discussion list)
> is strongly against it, and I believe the people who
> matter - the ones who work on the kernel and libc - are generally
> against it.
Right. Did the people on that list come up with any general plan for
how GNU/Linux vendors should transition?
I suppose I should subscribe to that list...
> I'd been interpreting this part of the policy amendment as saying
> "You shouldn't have filenames in packages (or created by packages)
> in non-UTF-8 encodings."
Well, that's not quite right. For filenames included directly in Debian
packages, or created by maintainer scripts, my policy proposal says they
*must* be UTF-8. For files simply created by running programs, it is
just suggested that they be UTF-8, for now.
> (I'm not generally a fan of
> filenames in non-ASCII UTF-8, but at least it's consistent.) If
> we're talking about what programs output, it should use whatever
> name and encoding the user asks for. We can't dictate what
> encoding end-users use; just what Debian packages use internally.
Are you saying that programs should attempt to convert filenames back
into the user's locale encoding in the actual filesystem, or just that
they should recode them for output?
Reply to: