From mboxrd@z Thu Jan  1 00:00:00 1970
From: reed@reedmedia.net (Jeremy C. Reed)
Date: Wed, 21 Mar 2018 21:25:10 -0500 (CDT)
Subject: [TUHS] syslog (was Re:  daemons are not to be exorcised)
In-Reply-To: <20180322002246.GK9739@mcvoy.com>
References: <CAFCBnZsS8Si6YGD4w_3W_znf3J5nVZGThkHVuQm5nD+FFm0DOg@mail.gmail.com>
 <94366db0-293b-214a-23a3-c7c895e4d30b@spamtrap.tnetconsulting.net>
 <alpine.BSF.2.02.1803200023360.43136@frieza.hoshinet.org>
 <98451f10-9b1b-049b-61ed-fd73586572fd@spamtrap.tnetconsulting.net>
 <20180321023125.GC6850@thunk.org>
 <76E7D09E-023B-4A29-ACCD-AF9ED425EE5F@xs4all.nl>
 <e38e9935-8633-2cee-8cae-c5e8be180966@spamtrap.tnetconsulting.net>
 <20180322002246.GK9739@mcvoy.com>
Message-ID: <alpine.NEB.2.20.1803212017420.25928@t1.m.reedmedia.net>

On Wed, 21 Mar 2018, Larry McVoy wrote:

> On Wed, Mar 21, 2018 at 06:18:37PM -0600, Grant Taylor via TUHS wrote:
> > On 03/21/2018 03:16 AM, Jaap Akkerhuis wrote:
> > >I've been told that syslog was came in existence as a debugging aid for
> > >sendmai.
> > 
> > I can't prove to the contrary.  But that does seem a little extreme to me.
> 
> For what it is worth, the syslog/sendmail connection rings a tiny bell for
> me, I can't prove it either, but I feel like there was some connection.
> Is Eric on the list?

Allman's logging was somewhat in 4BSD (4.0).

#               -DLOG -- include log information.  This is probably
#                       only useful on systems that include the logger.

delivermail was build with that flag.
But it used log.h which was not shipped in the version I have.
Some of the code has ifdef LOG and some doesn't.

2.79BSD (from McKusick's disk) which is later has logmsg(3) manual and 
its corresponding syslog(8) daemon manual (both Feb 5 1981).

I cannot find those files online so here they are:
http://reedmedia.net/~reed/tmp-oicyi3t6984y/logmsg.3.txt
Notice that it says "12/31/79"

http://reedmedia.net/~reed/tmp-oicyi3t6984y/syslog.8.txt

But no code there for logmsg, initlog, nor syslog.

(Note that uucp's logent.c has a syslog() but that is different.)

McKusick said that syslog was first in 4.1c and official in 4.2.
In both places shipped with sendmail code.
He said it was one of the first applications to use sockets.

4.1c.1 version says "reads datagrams from an IPC port
(currently port 2222, for no good reason)"
It uses the /etc/syslog.conf file.

Allman's 4.1c.1 code says:
**      This program implements a system log, implemented as the
**      "/dev/log" mpx file on a pre-4.2bsd system or a port on
**      a post-4.2bsd system.

I think this code is not in the dspinellis/unix-history-repo
(if you find it, please teach me how to find it).

The 4.1c code entirely changed api naming and some usage from

       initlog("delivermail", 0, LOG_INDEP);

to

        openlog("sendmail", LOG_PID);

(even though comments in syslog.h and syslog.c still mentioned initlog)

and from:

logmsg(LOG_INFO, "%s->%s: %ld: %s", From.q_paddr, To, MsgSize, statmsg);

to like:

syslog(LOG_INFO, "%s: message-id=%s", e->e_id, p);

even though syslog.c there defined logmsg() use and  not syslog()

I may have overlooked but don't see any use of the code outside of 
sendmail in 4.1c. The logger.c  (without manpage) utility was included 
in the sendmail source.

4.2 shipped two version of syslog(3) source code that no longer had 
logmsg() api. (One in libc and one with sendmail.)

4.2 could started it with rc.local. While I didn't see other use of it, 
a syslog.3 showed another example use:

openlog("serverftp", LOG_PID);

4.3 finally had lots of use (including in contrib): comsat, courier, 
various mh daemons, nntpd, savecore, shutdown, lpd, telnetd, r services, 
inetd, named, and much more. Even the date tool:

        syslog(LOG_NOTICE, "set by %s", username);