[223] in athena10
Review of changes to Evolution
daemon@ATHENA.MIT.EDU (ghudson@MIT.EDU)
Fri May 30 12:47:06 2008
Date: Fri, 30 May 2008 12:46:21 -0400 (EDT)
From: ghudson@MIT.EDU
Message-Id: <200805301646.m4UGkLcw011891@outgoing.mit.edu>
To: athena10@mit.edu
I'm working on fleshing out the project entry for the feature "Users
can easily read MIT mail using Evolution or Pine." Our work on these
programs has been fairly extensive over the life of the Athena
project, so I'm breaking this out into an email discussion and not
just a blurb in the plan.
I reviewed all of the changes we've made to Evolution in Athena 9.x,
filtered out the ones which don't seem relevant (including bug fixes,
which I'm assuming have percolated upstream). The list is:
1. Make ~/.evolution private when first created
2. Pre-configure MIT mail settings on first invocation
3. Hesiod support
4. Cache messages in /var/tmp in preference to AFS homedirs
5. Adjust folder display order so MIT folders appear first
6. Avoid using ibex files since they are dependent on byte order
7. chmod attachments read-only before opening
8. Build with krb4 support
(1-3) can be handled in a wrapper script. Emulating Hesiod support in
a wrapper script is a little hairy, but adding the Hesiod patch and
associated build goo in the debathenificator framework is probably
hairier. Also, our Hesiod patch is a bit frightening. It is
unfortunate that username.mail.mit.edu does not work with Kerberos
authentication, especially since it would require only a small tweak
to fix that, but we can't rely on NIST making changes to the PO
servers before this fall.
(4) is a performance change. Now that the W20 network has been
upgraded it might be reasonable to let it lapse. It's certainly still
*dumb* to use a network fileserver to cache data from a network mail
store, but it might not be the kind of dumb we have to fix. The patch
is reasonably clean if we do need to still apply it.
I think we can let (5) lapse. It's not hard to see or identify the
MIT Mail folder list. If it's possible, we might want the wrapper
script to change the default name of the homedir mail store from "On
This Computer" to something more accurate for our environment.
I'm not sure if (6) is still an issue. We will only be supporting
little-endian machines, but there might be a new related issue due to
64-bit machines. For the moment I'm going to say we shoudl let this
lapse until we identify that
(7) has been adopted upstream, based on observed behavior under Ubuntu
Hardy. Yay.
(8) is a frustrating point. It requires a package modification to
address and it would be unnecessary if NIST would enable krb5
authentication on the PO servers, which in my opinion is long overdue.
Since none of the other changes require a package modification unless
(6) rears its head again, this may be the only reason we aren't able
to use the stock Evolution binary in Athena 10.
The resulting to-do items are:
1. Create a package named debathena-evolution-wrapper which handles
~/.evolution privacy, initial configuration, and Hesiod po server
lookup.
2. Create a debathenificator package for Evolution to enable krb4
support. (Hopefully the code is still there and it's just turned
off in Ubuntu.)
