Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!pasteur!helios.ee.lbl.gov!nosc!peanuts.nosc.mil!dennis
From: dennis@peanuts.nosc.mil (Dennis Cottel)
Newsgroups: comp.sys.apollo
Subject: Re: unix mail configuration in an internet of rings
Message-ID: <1081@nosc.NOSC.MIL>
Date: 2 May 89 15:00:02 GMT
References: <1537@lgnp1.LS.COM> <42f7c091.1b147@apollo.COM>
Sender: nobody@nosc.NOSC.MIL
Reply-To: dennis@peanuts.nosc.mil.UUCP (Dennis Cottel)
Organization: Naval Ocean Systems Center, San Diego
Lines: 24

In article <42f7c091.1b147@apollo.COM> nazgul@apollo.COM (Kee Hinckley) writes:
>... the version of DPSS used internally supports store and forward,
>so if the network is down the mail still gets sent eventually.

We use Berkeley Mail and sendmail with all nodes linked to a single
/usr/spool directory.  One problem with this: during the time the node
with this directory is down, users can't even launch messages with
Mail.

This is currently a general problem with spoolers on the network, also
including PRF (at SR9 anyway).  If the spooling disk is unavailable,
the user can't complete the task and go on to something else.  I would
think that what you want to do is to spool the job onto the boot disk
(which is guaranteed to be available since you are running on it), and
then have a daemon process collect and process it.

The only way around this that I can see for the mail issue is to run a
sendmail daemon (and /usr/spool directory) on every node.  I have
resisted this because (a) it seems like more admin work, (b) it's nice
to have a single address for all the Apollo users, and (c) Apollo told
me to do it the other way ;-).  Maybe I should rethink this.

	Dennis Cottel  Naval Ocean Systems Center, San Diego, CA  92152
	(619) 553-1645      dennis@nosc.MIL      sdcsvax!noscvax!dennis