Re: Mailing list addition of resent headers

[I accidentally sent this directly to Charles. Here it is for the 
rest of the list:]
On 5/22/01 at 9:28 AM +0100, Charles Lindsey wrote:

> OK, so if I go on vacation, and instruct my machine to send my mail 
> to myself(_at_)some-other-domain (I think the sendmail /etc/aliases file 
> would do it) then the Resent-* fields could be used.
Well, maybe, but I'm pretty sure no. Ask yourself: What address would 
go into Resent-From:? Who's doing the resending?
Resent-* fields are really intended for mail that has come out of the 
transport system and is then re-introduced back into the transport 
system. That's why they have to be marked with a new originator 
address (Resent-From), originator date (Resent-Date), etc.
What I think you're doing when you use /etc/aliases or a .forward 
file is to simply rewrite the envelope address and send the mail on 
to someplace else without actually taking that message of the 
transport system. That is to say, you're doing a 2821 operation, not 
a 2822 operation. (For example, you're probably *not* going to put on 
the Return-Path field at this point; you're going to wait until final 
delivery.)
>  Presumably each message would also acquire additional Received 
> headers in the process.
This probably confirms my above suspicion: Adding Received fields is 
something that is done in the transport system (i.e., a 2821 
operation). Yes, when you resend and add Resent-* fields, you then 
introduce the message back into the transport system and it's going 
to add Received fields. However, adding Received fields is not part 
of "the process" of resending.
The sematics of the Resent-* fields are: "I 
(Resent-From/Resent-Sender) got this message and I am resending it to 
you (Resent-To). I resent it at such-and-so-time (Resent-Date) and I 
also sent copies of it some other people (Resent-Cc), even some 
people I may not be telling you about (Resent-Bcc). An identifier for 
the resending of this message is blah-blah-blah (Resent-Message-ID). 
The message itself was sent by X (From) to Y (To) and A, B, & C 
(Cc/Bcc) at some-time (Date)." If you can't figure out who the "I" 
and the "you" are, chances are Resent-* fields are not appropriate.
> So far so good, but I still don't see why mailing list expanders are 
> forbidden to do this. After all, they do not alter the message in 
> anyway. There is no "Fwd" in the Subject; the From and To remain the 
> same (just the envelope that is altered for each recipient).
Right, it is just the envelope that is changing, and its changing 
more-or-less mid-transport. Again, since the message isn't being 
taken and out of the transport system and then put back in, it's 
probably not appropriate to use Resent-* fields in this case.
> Certainly I see nothing in RFC2822 to imply this is wrong.
2822 is purposely silent on the issue of mailing lists in general 
because they do all sorts of things that don't fit the standard 
model. (For example, they don't seem to take things out of the 
transport system, yet they introduce things like List-* fields and 
other stuff into the message body.) I agree with Keith that Resent-* 
fields are more appropriate for a user operation and mailing lists 
should do something different if they want to add identifying 
information to message headers.
pr
--
Pete Resnick <mailto:presnick(_at_)qualcomm(_dot_)com>
QUALCOMM Incorporated - Direct phone: (858)651-4478, Fax: (858)651-1102