2016-01-05 15:26 GMT+01:00, Rainer Weikusat <rainerweikusat@???>:
> It's certainly possible to program something like this but even the
> 'mount /usr in initramfs' Debian-text admits that there's presently
> nothing which would need this, just something 'someone' might create
> in future. Within the envisioned 'release goal', the only practical
> effect is thus to break systems using /usr on a distinctive partition
> but no initramfs ("Know them by their fruits"?), or at least, the text
> claims this. But the UNIX(*)-filesystem namespace is supposed to be
> device-independent and in absence of the special case of 'software
> needed to boot the system', no two directories are required to reside on
> the same physical device. That's a fundamental property of the system
> which exists completely indepedently of someone's inability (or
> unwillingness) to imagine of something this could be good for.

Breaking normal use-cases to favour a few use-cases (be it opinions
or that udev works only if /usr is already mounted) is considered
best practice in Linux development.

> I've actually used PAM for transparently migrating a flat-file based
> multi-user 'workplace server' to Kerberos which came in very handy when
> (after the death of the last 'real' X-terminal) the complete
> installation was migrated to SunRays. But that was more than ten years
> ago and a part of the only 'volunteering'[*] task I'm ever going to do.

It seems the freedesktop tendency is to have a good compromise
between windows and fully network-served desktop, while pushing
to ship a product that can serve VMs, sessions, whatever, through the net.
I'd like to give a damn about redhat wishing to sell an
all-encompassing stateless UX solution.
If it wasn't for sane systems like slackware, void or devuan, this
could end with a lot of new or redent windows, macos and openbsd users
and an OS without any decent userspace.
Linux is the new Motif afaik.