[conspire] Partitioning revisited briefly

[Rick Moen]

Quoting Nick Moffitt (nick at zork.net):

> I dismissed only people who carved up partitions for historical reasons.
> I mentioned the valid uses of trying out multiple distributions and of
> using filesystems more prone to corruption.  The dismissal of the
> multi-distribution case has to be read in the context of the
> conversation: I was editing documentation tailored to one supported
> distribution, and which presented the multi-partition carve-up as a
> rather distracting section in an otherwise easy-to-follow tutorial.
> [...]

I actually owe you a small apology for being cranky on that point,
because I have to admit that I didn't understand the context of your
remarks, and that they addressed a situation elsewhere -- which in 
retrospect is pretty obvious.

> I heartily second this.  The FHS description has managed to remain that
> rarest of gem: a standards document that ordinary people can follow.

It is, albeit it's been getting a bit bureaucratic and bloat-ey.  An
outline version would be nice.  (As one of the people guilty of
perpetrating 'How to Ask Questions the Smart Way' on the world, I know
how that can happen.)

 
> Consider also the effect of *absolute* path symlinks when a volume is
> mounted in a new location in a tree.

Good point.  I am willing to have /opt link to an absolute pathspec
because it's highly unlikely to need to be functional in a changed-root
environment (or, on my system, need to be functional at all), but it's a
bad habit.

In some cases, symlinking /usr/local/src/foo/bar to, say, /root/bar is
just less awful than symlinking to ../../../root/bar (as you say) -- but
../usr/local/opt/ as a target is certainly more than tolerable.