Further ideas to avoid looking at more than one copy of the same
mail sent to multiple email exploders:
1. the MTA should look at all headers in the to: and cc: fields
and if there is duplicates, then only send it once.
This has some interesting consequences. Let's say I want to send
a note to this list but to keep you from getting a copy. I compose
a message with both ietf-822(_at_)imc(_dot_)org and keld(_at_)dkuug(_dot_)dk in
the
To field, but with only one recipient in the envelope:
ietf-822(_at_)imc(_dot_)org(_dot_)
The list gets the message, looks at the To: line, decides that you
are getting a copy of the message by another path, and so doesn't
send you a copy. Since I didn't send you a copy either, you don't
get one.
2. The MUA could look at messages and remove those occurring
more than once.
This more-or-less works, but implementation is tricky.
A simple comparison of message-ids does not suffice; it creates
an interesting security hole. Again assume I want to send a message
to the list, but I don't want you to receive it. So I create two
different messages with the same message-id. One of them I send to
you. The other one I send to the list. When your MUA gets the
second message, it discards it, because it's already seen a message
with that message-id.
So you want something that compares not just message-ids, but
messages. (message-ids can be used as a key to find potential
duplicates) The top few received headers will of course be different,
so remove them from the comparison. Also remove headers commonly
added by lists like Precedence and Errors-To. If none of the message
paths munged the content, the UA can easily detect duplicates. But if
one path munged the reply-to field, or rewrote
<moore(_at_)cs(_dot_)utk(_dot_)edu> to "moore"
<moore(_at_)cs(_dot_)utk(_dot_)edu> or
changed 8bit mime to 7bit, or changed boundary markers, or changed
content-transfer-encodings, or translated character sets, or converted
plain text to html (yeech), it becomes more difficult to recognize
two messages as the same.
For my personal message store, a simple comparison (not counting a
few well-chosen top-level headers) seems to work quite well.
But this would vary considerably from one user to another. It would
not work for lists that do extensive header munging. ( other
approaches at duplicate suppression also fail in the presence of
some kinds of list munging.)
Keith