Re: [Gnus 5.2.14] message-make-sender

[Sudish Joseph <sudish@ix.netcom.com>]

Richard Pieri <ratinox@unilab.dfci.harvard.edu> writes:
>      authentic   =   "From"       ":"   mailbox  ; Single author
>                  / ( "Sender"     ":"   mailbox  ; Actual submittor
>                      "From"       ":" 1#mailbox) ; Multiple authors
>                                                  ;  or not sender
> 
>         The Sender mailbox  specification  includes  a  word  sequence
>         which  must correspond to a specific agent (i.e., a human user
>         or a computer program) rather than a standard  address.   This

The "standard address" referred to here is one set by a user (for the
From: header, for instance); it does not mean just "valid RFC822
address" in this context--though all standard adresses must be valid
RFC822 addresses, of course. :)

>         expected to be a computer system term, and not (for example) a
>         generalized person reference which can  be  used  outside  the
>         network text message context.

A "generalized person reference" here is an intended-for-humans
address or name.  

> [In fact, not only does Sender: not need to contain a valid Internet
> mail address, it generally should not contain one. --Rat]

This is incorrect.  If you look at the ABNF above, you'll note that
both "from" and "sender" require the use of a "mailbox" (sure, sender
is optional, but if present it MUST have a "mailbox").  "mailbox" =>
"addr-spec" => "domain" (with other stuff, of course).

> aj> It is used when mailbox of From field is not senders real address or
> aj> it contains multiple mailboxes.

This is correct.  It tries to indicate the originator in ambiguous
situations like the above.

> No, that is the Reply-To: header.

No, the reply-to: address is a user-settable address, just as from: is
(both are "standard address"es in the context of the first quoted
paragraph).  Sender is there for one and only one reason: a lame,
half-hearted attempt at authentication; which -can- be very useful,
see below.

To further clarify the important distinction between from: and
sender:, it's worth noting that the DRUMS WG (currently updating /
clarifying 821 and 822 (*)) seriously considered strongly recommending
replacing the current use of reply-to with the from header.  This was
inspired by the complete chaos that prevails w.r.t. mailing lists that
hijack reply-to for their own uses (some do, some don't).

I.e., the intent was to abandon reply-to and make people use the from
header to indicate where replies should go (every mailer I've seen
does this properly: it selects from as the default place to send
replies).  This is in sharp contrast to using it to indicate where it
originated.  In such a context, sender: would be very useful.

-Sudish 

(*): They seemed to be in agreement about throwing the (useless)
definition of references: in mail and replacing it with a definition
closer to that of 1023.  The funny thing about that mailing list was
that I've -never- seen another list where so few messages had
references headers.  Even my procmail setup, where I munge one from
in-reply-to if possible, was mostly useless.  I guess when you've been
using mail as long as most of those people have been, it's difficult
to switch to a newer reader. :-)