Re: [Exim] Should vacation messages go to reply_address or r…

Αρχική Σελίδα
Delete this message
Reply to this message
Συντάκτης: Exim Users Mailing List
Ημερομηνία:  
Προς: exim-users
Αντικείμενο: Re: [Exim] Should vacation messages go to reply_address or return_path?
[ On Saturday, August 12, 2000 at 17:42:18 (+0300), Vadim Vygonets wrote: ]
> Subject: Re: [Exim] Should vacation messages go to reply_address or return_path?
>
> But there's a problem with the mailing lists, which you indicate
> below. In particular, say that someone replies to the message I
> sent to a mailing list, sending the reply both to me and to the
> maling list (therefore triggering the "personal" condition in
> Exim), setting the Reply-To: header to the address of the list.
> The vacation message will be sent to the list in this case.


This is still not a problem for the vast majority of lists where a
"precedence" header will short-circuit the reply. Even the rogue
LISTSERV mailer is handled properly by my rewrite of *BSD vacation
(though not the original) by not responding to messages sent from any
"listserv" mailbox.

Not that other non-*BSD derived vacation-like programs get this right.
I do get dozens (in the summer, fewer at non-holiday periods) of
responses every time I post to a LISTSERV list. These are usually from
brain-dead client mailers or from some brain-dead enterprise mailers
like Lotus'. I tried putting my own "Precedence: bulk" header in a
posting and I think that worked better, but unfortunately attempts by to
convince L-soft to do this by default, even by their customers
forwarding all auto-responder messages to L-soft support, have not yet
been successfull. Of course they are long-time professionals at
misreading RFCs (witness LSMTP for NT)! ;-)

> In this case, they can never receive bounce messages. And if
> they cannot receive bounces, they've got a bigger problem than
> inability to receive vacation messages. In my opinion, vacation
> messages may be treated as bounce messages, and a sender should
> be equally willing to know about his mail not reaching the
> recipient and the recipient not being able to read his mail.


On the contrary, the *sender* will receive a bounce (unless the
originating site is really screwed up!), just as is intended.

I think you're view of RFC 822 has been too narrow -- you should re-read
it again with this in mind. There are many many possible ways in which
the reply address(es) will be different than the sender address.

The difference between vacation replies and bounces is in intent. RFC
822 very explicitly allows for the sender to be different than the reply
address(es) and where that is intended the proper recipient of a vaction
notice is the reply address(es) (except of course for mailing lists).
Your opinion does not allow for this very necessary feature to be
implemented! :-)

> And vacation messages should be sent from empty address <>, like
> bounce messages, so no mailing loops will be created in case that
> both the original sender and the original recipient set vacation
> messages. In short, I think that vacation messages are like
> bounce messages.


No, they shouldn't be sent with an empty sender address. They should be
sent as if from the mailbox they are responding for. You do not want to
confuse them with bounces from a mail transport.

*BSD vacation has always depended on the fact that vacation messages
will have a "Precedence:" header which will prevent it from replying to
its own messages, and more importantly those of other vacation instances
too!

Vacation programs probably shouldn't respond to any other kind of
autoresponder either. However with the exception of a mail transport,
all such avoidance must be through the de facto standard (contrary to
what the *informational* RFC 2076 claims!) "Precedence:" header, or
through the fact that the message is not directly addressed to the
mailbox being responded to (i.e. in a "to" or "cc" header). Mail
transport responses (aka bounces) are of course avoided by not
responding to messages from "MAILER-DAEMON" (also just a de facto
standard used to represent the null return path).

The only advantage to sending vacation reponses as if they are bounces
(i.e. with an empty envelope sender address) would be to avoid having
them bounce back to you. However since this is already handled properly
by detecting the "Precedence:" header, and since it's hopefully rare,
it's not really necessary to do so.

> Which is a clearly bad thing to do. If I want to reply to you
> personally to a message you sent to a list, I don't want my reply
> to be sent to the list.


No, on the contrary it's a very good thing to do on a discussion list! ;-)

I normally want you to send to the list because I'm participating in a
"public" discussion. If you want to go off topic and don't want to
reply to the list when you're replying to a message from the list,
especialy when the originator wants you to reply to the list, then
you're doing something special and you need to realise it so that you
can adjust the addresses as necessary before you hit the send key.

In any case it is generally always much better to simply start a new
message to me instead of pretending that your private message is a
continuation of the list thread....

-- 
                            Greg A. Woods


+1 416 218-0098      VE3TCP      <gwoods@???>      <robohack!woods>
Planix, Inc. <woods@???>; Secrets of the Weird <woods@???>