[16153] in Athena Bugs

home help back first fref pref prev next nref lref last post

Re: sun4 8.2.9 : telnetd & the release

daemon@ATHENA.MIT.EDU (Greg Hudson)
Thu Aug 13 15:56:32 1998

Date: Thu, 13 Aug 1998 15:56:30 -0400 (EDT)
From: Greg Hudson <ghudson@MIT.EDU>
To: Karen Walrath N1XUL <karen@MIT.EDU>
Cc: bugs@MIT.EDU
In-Reply-To: "[16135] in Athena Bugs"

> 	The release really shouldn't be changing options like this

Jonathon mostly answered this question (and the new "mkserv remote"
option should be adequate, I hope), but let me give just a little more
information.

There is a set of files (listed in /usr/athena/lib/update/configfiles)
which are of interest both to the Athena release and to the system
administrator.  We make a minimal effort to allow the administrator to
make changes to such files, but it's only so effective.  A patch
release will stomp on those files only if it has a change to make, and
during the update a message is displayed like (you can look in
/var/athena/update.log):

	The following configuration files have changed and will
	be replaced.  The old versions will be renamed to the
	same name, but with a .old extension.  For example,
	/etc/shells would be renamed to /etc/shells.old and a
	new version would take its place.

	        /etc/athena/inetd.conf

Obviously, for auto-update machines, this approach is suboptimal: the
administrator won't see the message quoted above, and they might not
even know that an update happened.

One option for administrators is to make a /var/server/.private script
which redoes changes to config files if necessary.  It will be run
automatically after each patch release is taken.  But it is, of
course, harder to script the editing of config files than it is to
just edit them, so this approach isn't optimal either (and I don't
know how well documented it is).

For the 8.3 release, I have some ideas about merging changes to
configuration files of sufficiently restricted form (inetd.conf being
quite tractable, for the most part), but the presence of mkserv makes
it a less trivial problem.  So I'm not promising anything, but we are
aware of the systematic problem and have ideas for making the
situation better.

home help back first fref pref prev next nref lref last post