I received plenty of replies. Thank you all for taking the time.
"This list is great" - a cliche, but true.
The consensus was:
*** Sendmail's remote mode is BROKEN. ***
(in many more ways than the ones I reported).
The solutions fell into three categories:
(1) Sendmail has more problems than just remote mode. Use IDA-sendmail
instead. (I don't doubt this would have worked, but I have an
aversion to installing unofficial software in place of the Sun
stuff when the latter can be made to work. Maybe Sun's sendmail is
too broken for this to be sensible, but anyway...)
(2) Change /etc/sendmail.cf on clients so it forwards all mail to the
mailhosts explicitly, without using remote mode.
(3) On the non-mailhosts, replace /usr/lib/sendmail with
(owner=root, perms 755 or 555 -- NOT suid)
#! /bin/sh
rsh mailhost "/usr/lib/sendmail $*"
This should cause all sendmail processes to run on the mailhost.
I sort-of feel I should be using (2), but I tried one subsidiary
sendmail.cf that someone was kind enough to send me, and it solved all
the problems I'd mentioned...but no longer allowed mail to be sent
from clients to remote sites! I tried another set of suggestions for
modifications, but that left me with other problems. Given that I'm
happy with shell programming but regard Sendmailese as only slightly
less difficult than Hungarian, I decided to go for a modified version
of (3):
* mv /usr/lib/sendmail /usr/lib/sendmail.real
* Remove /etc/sendmail.cf on all but the mailhost
* Install this script as /usr/lib/sendmail (note the use of $@ rather
than $*, to deal with quoted arguments):
----
#! /bin/sh
if [ -f /etc/sendmail.cf ]
then
/usr/lib/sendmail.real $@
else
rsh mailhost "/usr/lib/sendmail.real $@"
fi
----
* Modify /etc/rc.local so that it starts /usr/lib/sendmail.real,
specifying -C/usr/lib/sendmail.subsidiary.cf if /etc/sendmail.cf
doesn't exist. This last is to deal with the (hopefully rare)
situation that someone sends mail to one of our non-mailhosts.
I will regard this question as closed, but if anyone wishes to berate
me for the nonstandard nature of this solution, I will be more than
happy to listen, especially since solutions (1) and (2) look sensible
too.
Thanks to:
Perry_Hutchison.Portland@xerox.com (who suggested solution (3))
and also:
Neil W Rickert <rickert@cs.niu.edu>
Richard.Hellier <rlh@cppuk.co.uk>
Thomas Weihrich <Thomas.Weihrich@arbi.informatik.uni-oldenburg.de>
Christopher Davis <ckd@eff.org>
per@erix.ericsson.se (Per Hedeland)
Robert L Krawitz <rlk@Think.COM>
ups!basil!pauline@fourx.Aus.Sun.COM (Pauline van Winsen)
Claus Assmann <ca@mine.informatik.uni-kiel.dbp.de>
Larry Chin <larry@cchtor.cch.com>
Greg Kulosa <gkulosa@oberon.com>
ccrwest!rich@UCSD.EDU
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:28 CDT