[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Zmail over DNA

    Date: Wed, 13 Apr 88 05:35:48 pdt
    From: ragosa%psyche@BOEING.COM (Rich Ragosa - BH AI Center - (215) 591-4295)

    Our lab has several Symbolics machines (1 3670 and 4 3620's) tied to a
    VAX 11/750 via DNA.  We use the VAX as a file server and Email router,
    but it would be very nice to be able to send messages from Zmail
    directly to users on the VAX and to machines tied to the VAX. This may
    be trivial, but is there a hack that allows this kind of activity?

Hmm, lemme see now, where to begin.... (and it's real late too)

I don't know about 7.2, but in 7.1 there are several bugs with the way
DNA uses/abuses your mailbox.  In short, while mail to/from the
Symbolics and VMS is possible, it is not robust.  The users will always
have to be be on guard.  If you have TCP for VMS you are advised to use
that instead of DNA.

First thing I did was to set my VMS mailbox forwarding address to my
machine:

MAIL> SET FORWARD MARLOW::JERRY

This didn't work.  When I received lots of messages, the DNA mailer
software croaked and left my Symbolics mailbox "busy" and dead.

Second thing I did was to set my VMS mailbox forwarding address to our
Symbolics Mailer machine.

MAIL> SET FORWARD MAILER::JERRY

Then I set the forwarding tables on MAILER to forward mail to me at
MARLOWE.

This sort of work, but it didn't.  I no longer had any problems with a
busy mailbox, DNA and the mailer work much better than DNA and DNA-MAIL.
But, whenever MAILER was down for other reasons, this hung every mailing
list on VMS.  That is, since VMS does not have a store and forward
mailer, people using VMS mail to send a message to me or to a mailing
list I was on got trapped and told that MAILER was not reachable.  This
confused almost everyone and was certainly very annoying.  I have know
idea how VMS users can live with such cruft.

Nevertheless, I use my Symbolics EXCLUSIVELY to read my mail, and ALMOST
EXCLUSIVELY to send it.

What I have done is to buy the PMDF package for VMS which provides a
"Pascal Memo Distribution Facility" ala MMDF from Innosoft.  They can be
reached at Harvey Mudd College by calling (714) 621-8006.  Ask for Ned
Freed or Kristin Hubner or send mail to ned@ymir.bitnet.  The package
costs $100 and provides for a VMS mail gateway to bitnet, many TCP smtp 
implementations, MMDF or PMDF over serial lines, and yes, even DECNET
and DNA.  It is extensible to provide support for an arbitrary SMTP
link.  It also provides a store and forward mailer for VMS as well
as an RFC 822 mail/domain mechanism.  Above all, and unlike many other
products, it uses standard VMS mailboxes which means that it is almost
invisible to other VMS lusers.  (They do have to read the incredibly
long and gory mail headers which all real mailers hide.)

I still use a mailer machine and not my own, because the busy mailbox
bug still exists.  But now, when the mailer machine is down, PMDF queues
it up and tries it again later.  Yay!

So, I set up a Symbolics Mailer machine and stuck it up on DECNET.  I
played with the mail forwarding files to forward mail to my own machine.

Next, I went onto VMS and set my mailbox forwarding address to forward
all of my mail to me at the Symbolics mailer machine.

MAIL> SET FORWARD IN%"'MAILER::JERRY'@V3"

(The above is not correct, you need alot more quotes, but you need to
get it to look like that!)

And things work.  In fact, the Symbolics mailer with DNA passes messages
from the VAX to the rest of the machines here, Suns and Symbolics.

We now have TCP/IP for VMS from CMU ($100 and it only breaks once a day)
and if we can get a Pascal compiler for VMS, we should be able implement
a much better mail service by eliminating DECNET.  Note, while the
Symbolics DNA has it's bugs, the real problems are mail the DECNET way
as opposed to the rest of the world.

I will be happy to provide more details should you actually buy PMDF,
but as for me, I am moving to 7.2 as soon as I can1 0(in which I am told
my DNA mail bugs will be fixed and also where the Symbolics domain
server understands MX forwarding records (I'll believe that when I see
it)), and getting a VMS Pascal to hook me up to TCP with VMS.

Anyway, why don't I send all my mail from the Symbolics?

Several reasons:  there is a known bug where three VMS recipients crash
the thing, and there is the problem in which you can't send mail to a
VMS mailing list (read logical name).

In addition, I have a few problems with Symbolics mail myself, which you
may wish to consider.

1)  Unlike every other mail system I have ever been on -- Unix, VMS,
    Multics, ... - the symbolics mail system has no understanding of
    aliases, that is, some shorthand symbol for some long gross mail address.

2)  The addresses automatically generated to reply to a message are
    certainly clever, but most often wrong.  Now it certainly seems as
    though one can modify these things using some kind of template
    mechanism, but there is nil documentation on how to do it.

3)  The mailer software is incredibly hard to use.  The postmaster has
    no control over what happens.  The postmaster cannot delete
    messages in queue, cannot read messages in queue, cannot move
    messages from one queue to another; in short, cannot perform any of
    the mindless tasks that always occur in a mailer environment.

    I have some patches that help the situation, but it is still pretty
    bogus.

4)  Due to the format of a mailbox, it takes forever and a day to read
    mailboxes in.  I'm sorry, that is wrong.  Forever and two days.


Jerry Bakin
Inference Corp.
(213) 417-7997
<@Riverside.scrc.symbolics.com:jerry@inference.dialnet.symbolics.com>