[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: / -> /usr symlink

On 15 Jan 1999, Thomas Bushnell, BSG wrote:

> "M.C. Vernon" <mcv21@cus.cam.ac.uk> writes:
> > I disagree here (nb - I am not a compsci). I think to begin with we should
> > have the /usr dichotamy, with the intention to change it later, when hurd
> > no longer has to worry so much about being linux compatible.
> I think the problem with this could best be explained by reference to
> history:
> /usr goes back at least to V6 Unix, and it was the place to put user
> home directories.  At some point some large binaries started to be
> popular, and /usr/bin showed up as a temporary hack, with the
> expectation that the root filesystem would get expanded.
> You can see how well the "change it later" strategy worked there.

And /usr/bin is such a terrible cludge. It impairs performance terribly,
and makes the sysadmins job a nightmane </sarcasm>

Seriously though - It might have been intended to change it, but I don't
personally see it as a practicle problem.
> I think we should patch the small number of packages that have a real
> problem, and move on.  We can create Hurd-specific patches for them,
> and submit them for consideration in general.  I'm happy if the
> general packages just did [ ! -l /usr ] or whatever.  I don't have any
> interest in trying to tell the ae package that they should not create
> /usr/bin/vi, but I do think we could say "please don't create
> /usr/bin/vi on Debian GNU/Hurd" and make it with a suitably generic
> test.

I don't think you have grounds to say "small"[1]. Part of the problem is
that we (as hurd developers) do not have the weight (i.e. a production
system) to throuw around to try and get a number of linux packages patched
so they cross-compile neatly when our only justification is philosophy. I
see the future of hurd as moving away from unix/linux, but not just yet.
To being with, we must try and get lots of stuff ported, and implement all
the desirable features in hurd (I want to re-implement mach as well, but
that's another matter for some years off) first. Then we are in a position
to move on. It's just a case of making sure we do move on - otherwise hurd
will be 'just another linux'. It takes a bit of vision, and a lot of
discipline, to do things like remove the symlink, with the intention of
putting it back in the future, but I think it is the best way forwards.
Gently gently - we want to win linux developers over, not alienate them



[1]: there are 2000+debian packages. Have you checked them all? My system
only has 526 on it - about 25%
Elen sila lumenn' omentielvo

Steward of the Cambridge Tolkien Society
Selwyn College Computer Support

Reply to: