XDG_SESSION_TMPDIR

Lennart Poettering mzkqt at 0pointer.de
Sun Apr 25 06:52:47 PDT 2010


On Sun, 25.04.10 10:43, Oswald Buddenhagen (ossi at kde.org) wrote:

> > Also, there already is XDG_CACHE_HOME, which is better aligned with
> > what traditional /tmp does. Your XDG_SESSION_TMPDIR kinda overlaps
> > with that, doesn't it?
> > 
> nope. caching is expected to outlive sessions and reboots. the spec
> doesn't mention it anywhere, but it is most certainly meant to
> correspond with the FHS's definitions.

Well, it says "non-essential (cached)". For me that reads as if it can
be dropped without any bad effects.

> > > The directory is always on a native local (read: not NFS or FUSE)
> >
> why no FUSE, anyway? why is "native" a hard requirement, not merely a
> strong suggestion?

Simply because all those weirdo file systems have issues with
guaranteeing atomic renames, AF_UNIX, POSIX locking, correct mmaping(),
character sets, case insensitivity, name/path lengths, correct handling
of sparse files, xattrs, symlinks, hard links, accuracy of timestamps,
lack of some of atime/mtime/ctime, stable inode numbers, correct
inotify, fpathconf() validity.

Normal file systems, such as ext3 or tmpfs, actually provide all of that
correctly (well, tmpfs doesn't do xattrs, but otherwise it does
everything right).

NFS normally has issues with atomic renames, with AF_UNIX and POSIX
locking, with mmap(), with xattrs, and inotify. In some setups those
features might exist, but more often than not they are broken. Ask the
gconf folks about it!

Lennart

-- 
Lennart Poettering                        Red Hat, Inc.
lennart [at] poettering [dot] net
http://0pointer.net/lennart/           GnuPG 0x1A015CC4


More information about the xdg mailing list