Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!SPEEDMETAL.ENGIN.UMICH.EDU!paul
From: paul@SPEEDMETAL.ENGIN.UMICH.EDU (paul killey)
Newsgroups: comp.sys.apollo
Subject: Re: Unix mail problem
Message-ID: <8910091730.AA00689@speedmetal.engin.umich.edu>
Date: 9 Oct 89 17:30:35 GMT
References: <30030@pbhya.PacBell.COM>
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 40


	We have about 460 users on 86 nodes. All of the Unix mail is located on one
	node (//s5/sys5/usr/mail).

% wc -l /etc/passwd
    6684 /etc/passwd
%
% ls -l /usr/spool/mail
lrwxrwxrwx   1 root           28 Sep 29 23:43 /usr/spool/mail -> //mail/bsd4.3/usr/spool/mail
% ls -lL /usr/spool/mail | wc
    4067   32530  214157
%

We run sendmail.5.59 plus some changes on the apollos.  5.61 with the
same kind of changes applied will be installed pretty soon, I hope.

The nature of the changes are for concurrency control when doing file
i/o and in dealing with temporary failures when verifying addresses
(rgy down, alias file not available, etc.).

I changed sendmail, binmail, /usr/ucb/mail, and MH to co-operate in
this endeavor.

I also made a couple changes to dbm, and have sendmail not using
dbminit(), but dbm_open() instead.

My colleague and unindicted mail co-conspirator, Mark Giuffrida, has
made every effort to get our code sprung out, but I have turned out to
be quite an immovable object in this regard.  But a current project is
to get sendmail on all our platforms (Sun, Dec, apollo) to compile and
run from the same set of source, and I will make that available.  I'm
*really* almost done!

Thanks to some work at Berkeley, other source code has been declared to
be free of AT&T license restrictions (like binmail, I believe) and I
assume that will make it easier for people to deal with things.  I
understand a freely distributable version of dbm routines is coming
down the pike as well (not from Berkeley).

--paul