For the first time in my scattered periods of dabbling in list maintenance
I'm running one on hardware that uses true sendmail instead of Smail or some
other substitute.
My misconception about owner-listname aliases was that they were invented for
NDNs on mail *from* the list -- and indeed they are very commonly used that
way -- and that some MTAs receiving distributions for subscribers might
somehow sense that the item was a mailing list distribution and seek to send
an NDN (for failure to deliver to a subscriber address on their site) to
[EMAIL PROTECTED] regardless of the envelope sender address on a
list distribution, so it was a bad idea not to have an owner-listname alias.
[Stephen van den Berg advises in SmartList's documentation that owner-
aliases are incompatible with the package, and now I can guess why.]
The sendmail(8) man page has always stated something to the effect that if
alias1 and owner-alias1 are both valid aliases, NDNs for failure to deliver
*to* alias1 will go to owner-alias1 instead of the sender of the message.
I didn't quite grasp that until I ran into it. All these years of happy use
of owner-listname in the envelopes of outgoing distributions and I finally
learn what it is supposed to do, and I really dislike the way it does it.
You see, under sendmail-8.9.0 and SunOS 5.6, when a subscriber submits to the
list and the article is *successfully* delivered to the submission alias,
sendmail clobbers the original envelope sender on the UNIX From_ line with
the address to which owner-listname expands (or with owner-listname itself
if it expands to a pipe). The owner- magic doesn't kick in only after
failures to reach listname; it also does its red heifer act on perfectly
healthy deliveries.
Since newer sendmail versions do not add Return-Path: any more, the original
envelope sender information is lost. If it did add Return-Path:, I'm not
sure whether that header would retain the original envelope sender's address
either; Return-Path: might be added after owner-listname clobbering, or it
might also be clobbered, so it may just repeat <owner-listname> or its
expansion, same as the From_ line.
If this sounds as though I were about to digress into a sendmail question,
sorry. This truly is going to end up as a list management question.
I've found that a large contingent of my subscription base cannot for their
lives learn the difference between listname@site and listname-request@site.
Not only do they send subscription service requests to the submission address
but also they regularly send articles to -request. Thus I have an autore-
sponder on the -request address explaining how mail to it is handled, in-
cluding a reminder that if what the person sent was an article for the mem-
bership, he or she needs to remail it to the submission address. Some make
that mistake again and again and need to be told again and again, so there is
no cache of who has already seen the acknowledgment; it goes out in response
to every piece of mail the -request address receives except blind carbons.
Sometimes experienced subscribers have reasons to write to -request, and they
don't need to get yet another copy of the automated acknowledgment. They
know what they're doing and they did it right. So I've offered to them that
they could write to owner-listname and get mail to me without receiving
another copy of the acknowledgment. I have subscribers who write to
owner-listname frequently for that very reason.
Now, I dislike not being able to find out the envelope sender. It sometimes
uncovers reasons to trust or not to trust the source of a submission (I
moderate the list) as being who or what it says it is. Granted, many
envelope senders are forged addresses anyway, but I find that mostly on mail
to listname-request, which has been publicized and receives spam, rather than
on mail to listname.
Thus, I'm asking the list-managers list for advice or suggestions. I have
thought of some answers, and you folks can tell me the merits or drawbacks
of them and suggest others.
1. I could live with it. I'm not happy with that idea; it would be good to
learn the actual envelope sender of an article.
2. I could start to use listname-owner as the envelope sender of
distributions and to encourage advanced subscribers to write to me there
instead of owner-listname, and get rid of owner-listname when the last NDNs
have had time to come in for distributions sent before I made the change.
[That isn't very soon, because some sites keep retrying a locked-out ID or an
account that is over its quota for three weeks or even a month before they
give up and return failure to the sender.]
3. I could change the submission address from the traditional listname to
listname-post and turn the listname alias into an autoresponder. At first
listname would forward to the new address with an autoresponse to use the new
address in the future; later it would return the post with instructions to
submit it to the new address; eventually it would be removed as an alias and
mail to it would bounce. Problem: listname-post or whatever is very, very
non-standard ... but are there any standards any more? As long as
listname-request works to satisfy the RFC that was under discussion here in
July, does the submission address have to be just plain listname?
4. Before rushing into anything else, I could see if the sysadmin can
reconfigure sendmail to insert Return-Path: headers on incoming mail; maybe,
just maybe, Return-Path: can retain the original envelope sender address from
MAIL FROM:.
Comments? Suggestions? Advice? Thank you.
David Tamkin
