[4022] in Release_7.7_team
Plan for moving tetex to a locker
daemon@ATHENA.MIT.EDU (Greg Hudson)
Sun Sep 21 01:31:39 2003
Date: Sun, 21 Sep 2003 01:31:36 -0400
Message-Id: <200309210531.h8L5VawM001817@error-messages.mit.edu>
From: Greg Hudson <ghudson@MIT.EDU>
To: release-team@MIT.EDU
tex doesn't really fit in the release because:
* It's big, and most of the bulk is data files which can be shared
between platforms.
* It takes a long time to build, and a lot of the build time is
creating data files which can be shared between platforms.
* It has a big compressed tarfile of data files, which CVS can't
version well.
* It likes to put symlinks in both its source tarfile and its data
tarfile, and CVS can't deal with those.
* It likes to install utilities with generic-looking names like
"access" and "readlink", which the release is cleaner without.
* When we make big upgrades, we frequently break things, which means
leaving the old version around for users to run is good. We can
do that more easily with lockers.
* Lockers would allow us to more easily delegate maintenance to tex
experts who aren't members of our team, like we've done in the
past with the gnu locker. I don't think there is currently anyone
we'd want to delegate control to, but in the past there have been
people like amu who were both more motivated and more able to
maintain a tex installation than we were. That situation might
arise again in the future.
Here is my plan for the locker setup:
* We build and maintain lockers with names like tetex-2.0.4, using
RCS to maintain the sources.
* We seize control of the tex locker (currently owned by cfields and
bert) and fill it with attachandrun scripts pointing at
tetex-2.0.4. If we build a newer version of tetex or a different
distribution of tex, we can update those attachandrun scripts.
* For the next full release (9.3), we replace our tetex installation
with attachandrun scripts in the release pointing at the tex
locker.
I have prepared a tetex-2.0.4 locker as a proof of concept. I'm
guessing it should move to the athena cell before this plan goes live,
but that can be done with a volume dump since there are no AFS path
references in the locker.
I believe this plan will take negligible effort and will significantly
reduce our time spent maintaining the tex installation, as well as the
size of the release. Setting up the tetex-2.0.4 locker was a snap.