Re: I'm confused... where do X11 bins go?
Hi,
On Sun, May 30, 1999 at 10:24:38PM -0400, Alex Yukhimets wrote:
> > How so?
>
> Branden knows better. And he is already encouraged to overcome all of them :0
>
> > Oh come on. We change the pathes of almost any upstream packaging every day.
>
> Well, from /usr/local to /usr - I am cool with that.
And from /opt to /usr, as we did with KDE. And from /usr to /. And from
/usr/lib to /usr/share. And from /libexec to /sbin. What makes /usr/X11R6
different? Don't tell me it is large: Other software is, too. Don't tell me
it is tradition, I don't believe in tradition for the sake of it.
> > >, requiring change in a partition scheme of MANY
> > > existing systems,
> >
> > This is a point. But it shouldn't be too hard to find a different partition
> > scheme.
>
> It is not difficult to find. It is pretty inconvenient to actually
> repartition.
I already granted you this point. However, my /usr/X11R6 contains 45 MB of
data. What is the upper limit? Do people really live on such dangerous
edges that they can move 50 MB across partitions without repartitioning?
> > The Hurd has a symlink . -> /usr Did the earth change its rotation
> > direction? No. Did hell break loose? No. Did we have problems? Yes, exactly
> > two: /bin/vi as provided by ae and cpio symlink from /sbin to /usr/sbin.
> > What a big deal!
>
> Great. Play with Hurd. This is completely new system with near zero install
> base, not striving for compatibility with existing UNIX systems.
Please get your facts straight. The Hurd is POSIX compatible. Unfortunately,
most available softwrae isn't. I don't see what UNIX compatibility has to do
with the peculiars in directory layout. The creation of /usr is historical
baggage.
> Go ahead
> and change everything you want there. Noone would be hurt or annoyed.
> You have a clean slate - if you find something very usefull in this
> experimentation - I am sure it will find way "upstream" pretty soon.
What I found out which I think is very useful is that you can flat the
directory tree by linking usr to /. In the same way we can flat the
tree by linking X11R6 to .
> > We have grandfathered /usr/X11R6 long enough already. What actually is the
> > benefit? For major upgrades of the Window System or concurrent versions we
> > have our packaging system.
>
> Yes, we do. And it can handle multiple library version existing concurrently
> and in the same directory. We put up with it, but I still don't like it.
? What wrong with that? As long as the sonames are distinct.
> More serious drawback is when you would want to have two vesrions of the
> same binary linked against two different set of libraries (this may be
> required in case of the new version not being backward compatible).
> Then we will need to have different names for this binaries and handle default
> set by alternatives mechanism. To change default one would have to change a
> whole forest of the symlinks and default is on a per-system basis.
You could try LD_PRELOAD for a change.
Or rpath. Both work fine to specify exactly which sonames to use.
Use the correct tools and you will have less problems. We can't support any
broken administrative behaviour equally well.
> In case of different directories for binaried and libraries - every user can
> simply change PATH and live with default. Why change this convenient scheme?
So you are voting for putting everyting in /opt/package? Why have this
"convenient scheme" for X but not for any other large package? Why not for
Gnome and KDE (if the latter ever enters the archive). Why not for Emacs?
Two sort of measures don't scale well, you know.
And, if this administrative scheme is important for you, you can always roll
your own versions of X. Or you could work on making it possible to have
multiple versions of windowing systems available on the same system
concurrently, as it works with emacs for example.
Thanks,
Marcus
--
"The purpose of Free Software is Free Software.
The End and the Means are the same." -- Craig Sanders
Marcus Brinkmann <Marcus.Brinkmann@ruhr-uni-bochum.de>
Reply to: