Re: /usr/X11R6
Manoj Srivastava <srivasta@datasync.com> writes:
> Such a move would, in fact, break section 4.1 of the FSSNTD,
> and would also violate the FHS.
I think that's a little strong. As long as those directories existed
(as symlinks), I don't think you'd actually be violating the intent of
the FSSTND or FHS.
The reason this policy issue was raised in the first place is because
the FSSTND and FHS do not address the issue of what to do with X11
dependent applications.
> This proposal also does not have an ewasy way of transitioning
> between releases of the X WIndow system (like, release 7, or version
> 12. In the current method, one may have multple copies of the X
> window system on the machine simultaneously, keepin one the default,
> and allowing others to experiment with newer version at will. Also
> allows an easy roll back by changing 3 links.
It appears that X proper has reached the end-of-the-line (given the
Open Group licensing debacle). Anything that replaces it will
probably be a variant of some sort (ie. Y, Berlin,
XFree86-something-or-other). X11R6 is going to be with us for a long
time.
As somebody else mentioned, alternate versions could be installed into
an alternate location.
> Other reasons I can think of are compatibility with other
> Linux/UNIX systems, and backwards compatibility (replacing
> directories with symlinks maybe awkward with dpkg).
That's a good reason not to do it - it might be a big transistion.
Too big, perhaps.
On the other hand, we have the exact same problem (replacing a
directory with a symlink) that we need to tackle for the FHS
conversion. So if we were committed to doing this, it would make
sense to do it at the same time.
> In conclusion, I think this is a stunningly bad idea.
I'm not convinced it's the best idea either. I'm just trying to get
everybody to put their arguments on the table.
IMHO, the idea has some value, because it represents "cleaning up"
another filesystem mess. Debian is in the unique situation where we
could decide that moving /usr/X11R6 is the thing to do, and actually
have the wherewithall to do it. I doubt that any other Linux
distribution would be willing to spend time and resources on such
"cleaning up" activities.
Again, IMHO, it would be nicer if we put X on an equal footing with
all the other competing packages, and didn't give it special status.
But practical considerations, such as the amount of pain it would
create for the developers, may be too overwhelming.
Summarizing, we have four proposed policy options now:
a) /usr/X11R6 is for the X system only. All dependent applications (needs to
be defined) would be installed into /usr. Where the various libraries are
installed would also need to be determined.
b) /usr/X11R6 is for the X system, and all applications and libraries which
depend upon it. This option makes sense for people who want to NFS mount
/usr/X11R6, but not /usr (I'm not sure if anybody really does this).
It also segregates the apps so it is obvious what they are supposed to be
linked against when X11R7 (or some other succeeding system) comes out.
This may be useful in making a transition (ie. we could do pathname-based
ldso tricks like we did for the libc5 to libc6 transition). It seems
unlikely that X11R6 is going to be displaced by a successor anytime soon.
c) /usr/X11R6 is for the X system, and all applications and libraries which
decide to install themselves there by default. Other X dependent
applications and libraries may choose to go under /usr, at the whim of the
upstream author. In short, there would be no policy where X applications
should be installed. This is essentially the situation we have now. But
it should be written up, so that people won't file bugs about applications
being in the wrong place.
d) Everything gets installed under /usr, including the X system. Then
/usr/X11R6 would be a symlink to /usr. This has the benefit that we don't
have to make an arbitrary policy about what should and shouldn't go into
/usr/X11R6. Also, X would no longer be treated as a "special case". New
users wouldn't have to learn to look in so many places for the binaries
and other files. With the symlinks, we shouldn't have compatibility
problems.
The downside to doing this is that it is difficult to replace a directory
with a symlink - all the packages which install into /usr/X11R6 would need
to be changed at once. This is similar to the /usr/doc situation with
the FHS transition. Also, the other Linux distributions will probably
keep using /usr/X11R6, so this might be a controversial move. Many
packages would need to be changed, and everybody does not agree that this
is necessary or even useful.
Right now, we've got people advocating each of the four solutions. I
don't think we have come to any consensus agreements yet. A lot of
people like option d), including the X maintainer, but it is a bit
radical and some people are strongly opposed to it. I personally
don't like option c), because it's sort of a "cop out", but it is the
easiest thing to do. People are split between options a) and b) as
well. I'm pretty wishy-washy on my policy preference.
Cheers,
- Jim
Reply to: