GUMBY@mit-ai.arpa (David Vinayak Wallace) (11/18/86)
Date: Mon, 17 Nov 86 19:48:06 -0800
From: Marshall Rose <mrose at nrtc-gremlin.arpa>
there is a difference between working correctly and working. Working
correctly is to ignore sender and return-path. However this is not
often working...there is neither perfection nor diligence in the
Internet. It is better to know that you can fall back on sender
and/or return path in case you have to (when working on a poorly
constructed message), than to just say "you can't reply to that".
There's a difference between DWIM and incorrectness. Your mail handler
may reasonably complain "I can't reply to that" and ask the human for
help; it can't reasonably figure out which of From, Sender or Return-Path
might interest the user (for instance, if I want to send a message to
someone who will fix it up I might send a message to the Return-Path,
whereas in a legitimate reply I might be able to construct a valid
destination address).
For purposes of replying there isn't much difference between Return-Path,
Sender, and Subject. They all have inappropriate semantics. The machine
shouldn't fake out naive users by using random information.mrose@nrtc-gremlin.arpa (Marshall Rose) (11/18/86)
This is NOT, repeat NOT, a case of DWIM. The issue is simple: mail
composers should use "reply-to if" they want to redirect replies
that would normally go to the "from" field; mail repliers should
honor "reply-to" over "from", but need to be able to do something if
neither field is present. This is yet another example of the trite
"be conservative in what you send, liberal in what you accept".
This emphasizes the key point of my original message: the sender
can use "reply-to", but it is upto the receiver to do the right
thing. The right thing is to use "reply-to" or "from" for the
primary recipient(s) of the reply. There is no concrete right thing
for generating the list of secondary recipients (usually to and cc
are the right thing, at the user's option). For example, my
original message had a "reply-to" of header-people, and my mailbox
did NOT appear anywhere in the message except for "from". I assume
that Mark explicitly added my address to his reply. In this case
the mail composer used "reply-to" to redirect replies, and the mail
replier did not honor this protocol.
A final note: do not, as your previous message suggests, mistake
the operation of replying with other message-handling functions
(e.g., forwarding, and so on). If you want to send the message to
someone who will fix it up, you are NOT replying to the message.
You are sending a message with a different purpose, not a reply to
the original!
/mtrMRC%PANDA@sumex-aim.arpa (Mark Crispin) (11/18/86)
Marshall -
The Internet mail world is not yet such that it is impossible
to use a mail system that obeys RFC 822 when it comes to the
semantics of replying. The only reason for adding Return-Path to
your mail user agent's REPLY parser is laziness on the part of the
mail maintainer -- that is, he is unwilling to handle the problems
which occasionally come up and explain the realities of mail to a
user who thinks that the reply address is "always" available in a
Return-Path.
Various vendors then "validate" their mail software against
this lazy software, since it's located on a Unix and everybody is
running Unix so it must be right... Then the poor maintainers of
non-Unix software get long, pendantic complaint letters from the
above-mentioned vendors (or their users) stating everything imaginable
about the Internet protocols except the facts. Wollongong is a
worst case, but not the only case.
I don't want to argue English semantics, but "be conservative
in what you send" sure means to me that you should be conservative
and not send replies to the Return-Path. What's more, replying to
the Return-Path is explicitly forbidden by RFC 822.
-------mrose@nrtc-gremlin.arpa (Marshall Rose) (11/19/86)
Mark: If what we are arguing here is "validation", then I appologize. I am not talking about validation. I agree with your argument that validating against lazy software is wrong. /mtr
kre@munnari.oz (Robert Elz) (11/19/86)
This discussion is getting absurdly out of hand. No-one has suggested using the "Return-Path" for replies anytime that there's a From or Reply-To field in the header. It doesn't seem entirely unreasonable to me to allow replies to the Return-Path (or Sender) if there's no From in the message (which clearly indicates that the sending mailer is broken). But as reported, rfc822 forbids this. In that case, I think I'd just "fix" my mailer - if an incoming message arrives with no From: field, then copy the Return-Path to a From header... Same effect, and this isn't forbidden. The right solution is to fix the broken mailer - but that's someone else's broken mailer, and getting others to fix their broken programmes isn't always easy. But, the rfc822 comments on where replies should be sent cannot possibly be mandatory - its entirely irrational to try to forbid to whom anyone sends mail (at least just on the basis of what's in a header, ignoring political/financial problems). A reply is just a mail item, I can send it to anyone I like. All that's material here is for the headers to give some guidance on where replies are intended to go, and for mailers to use this guidance to establish the default recipients. Normally, if someone sends me mail with both From & Sender headers (indicating in the typical case that a secretary sent mail on behalf of someone else) I would send a reply to the From header. This would be a reply on some matter of substance in the mail. On the other hand, if I don't like the way the mail is typed (say I find some spelling error that offends me, or the layout makes the mail hard to read on my terminal - wider than 80 cols) then I think sending a "reply" to the Sender is entirely reasonable. Yes, that's not a "reply" in some narrow sense of the word. I intrepret the word more liberally though - any message I send that is a response, of any kind, to mail I receive is a reply (ie: if receiving the original mail was the stimulus necessary to send my mail, then my mail is a reply). Robert Elz kre%munnari.oz@seismo.css.gov