[2129] in athena10
Re: the impact of each metapackage
daemon@ATHENA.MIT.EDU (Tim Abbott)
Wed Apr 8 18:39:10 2009
Date: Wed, 8 Apr 2009 18:38:40 -0400 (EDT)
From: Tim Abbott <tabbott@MIT.EDU>
To: Jonathan Reed <jdreed@mit.edu>
cc: Debathena Trac <debathena@mit.edu>
In-Reply-To: <29CC3B52-34F4-45FF-A8BF-24775CC0A1E7@mit.edu>
Message-ID: <alpine.DEB.1.10.0904081608470.28854@vinegar-pot.mit.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
My concern with settings standards so that -standard messes with your
system and -clients doesn't is that I think most people running laptops or
servers want the current -standard functionality (in particular, most
people want AFS and the libraries needed to run most software in AFS).
Maybe I'm misunderstanding the intention, but I don't think that
advertising debathena-clients available as a "don't mess with my system"
alternative to debathena-standard means that we should not be very careful
to limit the complexity and potential for breakage caused by
debathena-standard, which I suspect will be the level of integration
desired by most users not using AFS home directories (i.e. on their
laptops or personal servers).
My general prefererence is that we try to ensure debathena-standard has
all the desired functionality without doing anything horrible. Coherent
options that we advertise at the top level have a significant cost in
terms of support, testing, and complexity; we should be careful when
considering creating more that they have a clear mission that there is
demand for. I think that for each of these options there is probably a
tradeoff between horribleness and importance; AFS can cause system
stability options, for example, but I think it is sufficiently important
that it belongs in -standard anyway.
I think it would be helpful to consider this broader issue in the context
of the individual proposals that have come up that effect it. The
questions that I can recall that spawned this discussion were:
(1) debathena-tex-config
(2) changing /etc/hosts on Debathena machines
I don't think that this is complete, though. Remind me what else there
is?
-Tim Abbott
On Tue, 7 Apr 2009, Jonathan Reed wrote:
> At release-team today, we discussed the various metapackages and the degree to
> which they should or should not affect the user's machine. There seems to be
> a desired for debathena-clients to lean towards the "make as few decisions for
> me as possible on my behalf". By the same token, debathena-workstation (the
> new one) is more along the lines of "make as many decisions for me as
> possible".
>
> Keeping this in mind, we thought about what configuration packages/diversions
> should be present in each package. We propose removing debathena-from-config
> and debathena-finger-config from -clients and putting them in -standard
> instead, thus making debathena-clients more along the lines of "Don't change
> things on my machine, only give me new Athena functionality". We propose that
> -zephyr-config, -kerberos-config, -hesiod-config, and -ldap-config stay in
> -clients, since those make packages we install more useful.
>
> We ran out of time and didn't consider whether anything else should be
> shuffled around between -standard, -login, -graphical-login, and -workstation,
> but I suspect there's not too much else we want to move.
>
> What do people think of this? (Other people who were at today's meeting
> should feel free to correct this if any of this is wrong/misleading).
>
> -Jon
