|
Searching XEmacs
Quick Links
About XEmacs
Getting XEmacs
Customizing XEmacs
Troubleshooting XEmacs
Developing XEmacs
|
|
|
Stephen Turnbull
<stephen@xemacs.org>
Volunteers eagerly sought!
We should reverse Steve Baur's decision to not distribute the SUMO
Lisp in the main tarball. The packages are popular, but the
typical user of the core tarball wants the kitchen sink.
This is not about reintegrating the packages into the core (a la
the GNU mainline). Nobody strongly supports that.
We're talking about making the default distribution include the
packages, or at least have an easily available all-in-one option.
We don't really have that now, even with the SUMO: you have to
fetch two tarballs, untar them in different places, make all these
choices, ....
People who balk at a 35MB tarball will still have the
disaggregated setup available.
Status
One proposal is as follows. Create the following series of
tarballs for each release.
- xemacs-21.4.0.tar.bz2
-
One-stop-shop. Unpack this under $srcdir, get all
core source and data, .elc, .info,
and the current SUMO tarball, in
$srcdir/etc/. (Maybe
$srcdir/xemacs-packages/, or
$srcdir/sumo-data/.
./configure; make works as always. make
install has a final step: if
$srcdir/etc/xemacs-sumo.tar.bz2
exists, unpack it into $PACKAGEROOT.
Baroque, yes, but remember those vi -using
sysadmins. They are used to being able to unpack into some
random tmpdir. Anyway, SUMO unpacking needs to
obey $datadir et al.
- xemacs-21.4.0-src.tar.bz2
-
Core source and data, can generate -elc and
-info. Connect cost optimization for the
weakly-connected.
Maybe this tarball should include
xemacs-pui-min.tar.bz2, and unpack it in
$PACKAGEROOT? Alternatively, put them in
$srcdir/xemacs-packages, to bootstrap
list-packages . Then the install process
automagically tries to invoke list-packages to
install pui-min packages in
$PACKAGEROOT, and only if that breaks use
install ?
- xemacs-21.4.0-elc.tar.bz2
-
Compiled Lisp, build time optimization for the well-connected.
- xemacs-21.4.0-info.tar.bz2
-
Generated Info, build time optimization for the
well-connected.
- xemacs-pui-min.tar.bz2
-
xemacs-base, efs: starter kit for
list-packages. Tiny subset of SUMO, with same release
schedule as SUMO. Also includes their dependencies, which I
believe is dired for efs.
- xemacs-sumo.tar.bz2 -> xemacs-sumo-2001-03-01.tar.bz2
-
Package SUMO, with release schedule separate from core. Steve
Youngs thinks Mule should be excluded.
- xemacs-sumo-mule.tar.bz2-> xemacs-sumo-mule-2001-03-01.tar.bz2
-
Package SUMO (Mule packages), with release schedule separate
from core.
Alternatives
Steve Youngs suggests ``instead of integrating all of the packages
back into the core, why not just xemacs-base,
mule-base, efs and perhaps
dired?''
Steve Turnbull replied ``I'd actually like to see more and more
stuff pushed out of the core (but that is controversial).
However, especially with well-maintained packages like dired and
EFS, I think it makes a lot of sense to keep them separate. And
surely the maintainer in question, Michael Sperber, favors
packages!!''
A better alternative for putting more capability into the core
might be Jan Vroonhof's proposal for a lightweight HTTP-based
specialized bootstrap package-fetcher in the core. This would
require that we set up some additional infrastructure on the web
site. There is no way we can rely on having this ready by March
1, unfortunately.
Is this really necessary?
Steve Youngs reports
Having my hard drive go up in smoke has meant that I have had to
re-install everything from scratch. When it came to XEmacs I
did ...
[...]
Everything fine, no problems at all.
But it is a FAQ; obviously somebody is having problems. One
important difference is that you knew what you wanted and had no
trouble with the ./configure and M-x
list-packages stages. Many people count on
xemacs.org to provide a clear recommendation for a
safe default install, and based on the FAQ count, we're falling
down on that.
I know, anybody who can find the AnyKey should be
able to install XEmacs with a little bit better docs. Of course
they can. The point is that there are two kinds of users (newbies
and busy admins) who could be better supported with some kind of
all-in-one distro.
Open bugs
None. (Well, of course the current situation is a bug.)
Other open issues
None.
Discussion
We should reverse Steve Baur's decision to not distribute the SUMO
Lisp in the main tarball. I don't know how much work this would
involve. It could be as little as a change to INSTALL ("we don't
support any means of installation that doesn't involve untarring
the SUMO in $prefix/lib/xemacs; if you want to avoid that, read
README.packages"). It could involve somehow including the SUMO in
the core distribution. Or something in between (don't ask me what
yet).
The packages are popular, but the typical user of the core tarball
wants the kitchen sink. The package system is used to keep
up-to-date on heavily used packages, or packages with important
new features. People who want a subset typically are either
capable implementing that themselves anyway, or should let their
Linux distro deal with the issue, or can remove unwanted packages
ex-post install via package-ui.
Closed bugs
None.
|