NetNews Usenet Archive 1992 #23/NN_1992_23.iso/spool/bit/listserv/maill/246

Internet Message Format | 1992-10-15 | 5.7 KB

Path: sparky!uunet!stanford.edu!lll-winken!bu.edu!wupost!newsfeed.rice.edu!rice!is.rice.edu!schafer
From: [e-mail] (Richard Alan Schafer)
Newsgroups: bit.listserv.mail-l
Subject: Re: questions on mail to bitnet nodes
Message-ID: <[e-mail]>
Date: 14 Oct 92 16:16:26 GMT
References: <[e-mail]>
Sender: [e-mail] (News)
Organization: Rice University
Lines: 105

In article <[e-mail]>, [e-mail] (Kurt
Carlson) writes:
|> BITnet messages from our system are posted from MAILER@ALASKA
|> direct to the recipient (not any mailer defined within XMAILER NAMES)
|> Class=M, Name=sender_userid, Type=Mail, with appropriate
|> RFC822 headers.

Why not?  By doing so, you're bypassing the place that those sites have
said they wish to receive mail.

|> Occassionally, users at our site foul up the recipient userid.
|> We've encountered several VM sites running RSCS v3 which are
|> sending non-delivery notifications as files back to MAILER@ALASKA
|> which is, needless to say, less than desirable.  My understanding
|> is that this is an RSCS v3 function and the original messages
|> never made it to their mailer as it was posted directly to the
|> user.

Yes, if you try to send mail directly to an invalid user, RSCS will
recognize this and complain.  Their mailer (as you say) never gets
involved at all.  (Another reason why what you're doing is a bad idea,
since the user may have been forwarding their mail using mailer functions.)

|> My questions:
|>
|>  1.  Must we (or even just "should we")  respect XMAILER NAMES and
|>      post bitnet messages to the registered mailer in bsmtp packaging?
|>
|> We have not done this to date for BITnet messages as we've not felt
|> obligated to and it is more efficient not to.  We do, of course,
|> package bsmtp and use DOMAIN NAMES registered mailers for gated
|> internet messages.

"Must we"?  No, probably no more than you're obligated to send
RFC822 format mail instead of VAXMail notes with :: things in the headers.
Must I send your users mail in English, or can I use Sanscrit if that's more
efficient for me to write?

"Should we"?  Absolutely.  Those sites had told the world "We are running
a post office function, please send mail to us there".  You're ignoring
their stated wishes, and denying them the ability to handle things like
mail forwarding that they have setup using their mailer.

"More efficient"?  Perhaps in some way it's more efficient for you to
generate the mail, but in the long run, the efficiency is certainly not
to ignore other sites configurations.  The problem that caused you to send
your note is an example of an "inefficiency" you're causing yourself by
having something which appears to be a mailer refusing to send to other
people's mailers.

Frankly that question almost comes down to: "Why should I care if the people
on the other end have trouble reading my mail, if it's harder for me to
send it in a fashion that they can read easily?"

BSMTP packaging is really just icing on the cake, although perhaps at
the root of your "efficiency" issue.  Although you *should* put BSMTP
envelopes around your mail if you can, I'd rather have mail be sent to
my mailer without a BSMTP envelope than have it sent to my userid directly.

|>  2.  Should sites in general accept direct posts irregardless of
|>      a registered mailer?
|>
|> We accept direct or to MAILER@ALASKA, all of which are passed through
|> our mailer which will bounce non-deliverables back to the recipient.

Yes, you should accept direct posts.  The guiding principle should always be
to get the mail delivered.  Of course, if accepting direct posts means
that the mail is unreadable by the user in any fashion, then I'd probably
change my mind.  But you clearly say that's not the case.  In fact, you're
taking direct posts and passing them to your mailer, probably taking advantage
of the way JNET handles NJE traffic to do so.  For other systems, that
may not be feasible.  Seems odd to me that if you think your mailer is so
important that you do this, that you wouldn't believe other sites mailers
are important enough to send to them, too.

|>  3.  In the event of non-deliverable messages, are there any
|>      standards on how they should be processed?
|>
|> Most sites, us and the RSCS v3 sites in questions, endeavor to process
|> them automatically without human intervention.

I'd have to go back and read RFC1123 and RFC821 to be sure exactly what
they say, but those documents do describe procedures for non-delivery.
Certainly, returning the mail to the originator of the file is a standard
action.  However, you should recognize that at the RSCS level, most systems
who haven't coded special handling of things will not distinguish between
a mail file and anything else.  That's probably why your mailer is getting
the bounce-backs, since your mailer is the NJE originator of the file.
Those sites run a mailer to deal with mail transport issues, which is yet
another reason why you should send mail to their mailer when you know about
it.

|> If there is some definitive CREN authorized document which answers these
|> questions, that would be my preference.  Lacking that, I'll listen to
|> whatever opinions are offered.

To be honest, I don't know if I've ever conceived of a site which ran a
mailer refusing to send mail to other sites mailers.  It seems basically
unsociable, and counter-productive to me.  Whether there is a CREN
document which specifically says that, I don't know.  A place to look
might be the NEWTAGS DESCRIPT document which describes the tags in BITEARN
NODES which define mailer information, but to be honest, I doubt that
it *requires* sites to send to those mailers if their sites has one.
--
Richard