🚀 go-pugleaf

<rusty.b.ramser@ksc.nasa.gov> wrote:
> The Problem (abridged version):
>   MMDF isn't delivering SMTP to any addresses that aren't
>   in my local domain area.
> 
> The Setup:
>   SCO Open Desktop UNIX v3.0.0 installed and running
>   TCP/IP installed and running
>   BIND (a.k.a. DNS) installed and running
>   MMDF installed and running incorrectly (?)
> 
> The Problem:
>   I selected to not configure a badhosts or badusers
>   channel -- the reasoning being, if MMDF properly uses DNS
>   there will not be any hostname that it can't resolve.
> 
>   Attempting
>   to mail to anyone outside of my domain (ksc.nasa.gov)
>   results in a failure message of the following type:
>      johndoe@company.com: (BHST) Unknown host/domain name
>      in "johndoe@company.com"

tls011.ltr talks about an error in the MMDF configuration files; the
default mkdev scripts work well when the badhosts channel is used but
not otherwise, in my experience.

When using a name server with MMDF, the SCO support database recommends
running `mkdev mmdf' and then changing the line
MTBL rootdom, file="root.dom", show="Root Domain", flags=route
to
MTBL rootdom, flags=ns, flags=domain, show="Root Domain"
and then doing `cd /usr/mmdf/table; su mmdf -c ./dbmbuild'.

There are other, different hacks which work; I personally have not
tried the above solution but rely on a more complicated setup.

One problem to watch out for when using a PPP connection in
demand-dialup mode (with SCO UNIX SVR3.2v4.2, TCP/IP 1.2.1, and the
net382c SLS patch dated 23 June 1994): the MMDF pseudo-sendmail does
not seem to bring the link up to contact name servers but instead times
out in about 5s when attempting to resolve the addresses.  The `real'
sendmail in the patch appears to force the connection up every time it
is run (even if only to examine an empty local mail queue), ensuring
that name servers are contactable before attempting to do lookups.
Hence using sendmail instead of MMDF for delivery of outgoing mail
works well.  Extrapolating, I suppose that another possible solution
would be to use MMDF but to replace the MMDF pseudo-sendmail binary
with a wrapper which pings the remote side of the PPP interface and
then calls the pseudo-sendmail, to guarantee the link being up.  The
/etc/if.ignore file does not seem to apply to any of the services being
used (if the manual page is correct), so that doesn't look relevant in
solving this problem.

It seems a lot easier to just ditch MMDF for outgoing mail and use the
sendmail supplied in the patch.  I can't attest to sendmail doing a
good job as an SMTP daemon, but as a delivery agent it seems to work
fine.  However, the standard setup doesn't set up the queue to be
processed periodically, so I had to add `sendmail -q60m' to the
startup sequence (a cron entry would be fine also).

Comments, anyone?
-- 
Andr\'as Salamon                   andras@is.co.za