Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site cadvax
Path: utzoo!linus!decvax!decwrl!amd!resonex!cae780!ubvax!megatest!cadvax!vahe
From: vahe@cadvax (Vahe Sarkissian)
Newsgroups: net.mail
Subject: Stale sendmail locks
Message-ID: <114@cadvax>
Date: Tue, 4-Sep-84 16:33:24 EDT
Article-I.D.: cadvax.114
Posted: Tue Sep  4 16:33:24 1984
Date-Received: Sun, 16-Sep-84 06:21:40 EDT
Organization: Megatest Corp., San Jose, CA
Lines: 29

I've encountered a problem with sendmail recently, for which the
fix appears straightforward.  But, I'm worried that there's something
I've missed, and am seeking advice.  Eric Allman, are you out there?

First, I run sendmail (Version 4.12) on 4.2 bsd on three 750's and
25 proprietary micros, connected via Ethernet.  The problem is that
I'll occasionally see a whole bunch of stale mail get delivered.
Reading the logs, I realized it was due to sendmail dying
(cause as yet unknown), and leaving a whole bunch of lock files.
When the machine reboots (not often; most machines here stay up for
about a month at a time), the locks are removed in /etc/rc,
and presto -- weeks-old mail at your doorstep.  In an environment
that is heavily dependent on mail for communication, this is unacceptable.

Well, the fix for stale locks is simple: time them out, with
some intelligent cutoff point.  But, as I read the documentation
(in the code and in the manuals -- well written, I might add),
I get the distinct impression that Allman knew about this, and
omitted it on purpose.  Other than the chance of mail getting
delivered twice, I can't see any other problems this could cause.

So, the QUESTION: am I missing something?  Is there some fundamental
reason why the lock files shouldn't be removed (presuming, of
course, that the decision is made in a reasonable fashion.)

I can be reached via:
	...allegra!fortune!megatest!vahe
	(408) 998-7110 x3168
	Vahe Sarkissian @ Megatest Corp, 880 Fox Lane, San Jose, CA 95132