From: Russ Allbery (rra@stanford.edu)
Date: Sat Mar 16 2002 - 17:06:57 CST
Bruce Lilly <blilly@erols.com> writes:
> Assuming for the moment that there exists some reason to forbid multiple
> msg-ids in Supersedes in Usenet articles, the question of what a
> mail-to-news gateway should do when it receives a message with multiple
> msg-ids in a Supersedes header remains.
The X.400 gateway meaning of Supersedes is considerably weaker than the
meaning of Supersedes in news. Inclusion of a Supersedes header in a news
message essentially makes it into a control message; it's a cancel message
in addition to a regular message, and can have significant destructive
properties considerably beyond what a Supersedes header can have in mail.
In short, the news header is not the same thing as the mail header due to
the nature of the medium. Changing the name of the news header is flatly
not an option without an intervening period supporting that header; it
predates RFC 2156 and is very widely implemented, regardless of whether it
was ever documented in an RFC. The news RFCs are hopelessly obsolete and
do not document the current protocol as used.
The proper thing to do with a Supersedes header in a mail to news gateway
in my opinion is to delete it, just like one normally does not allow mail
to news gateways to post control messages. Under special circumstances
where the target can be verified, it may be okay to let it through, but
I'd be reluctant to do so.
Adding a multiple message-ID form of Supersedes creates an ability to
delete multiple Usenet messages with a single message, an ability that is
not currently present on Usenet. It would do this without any functioning
authentication system, no mechanisms to prevent abuse, and no addressing
of any of the underlying security issues. As such, I oppose it for the
health of the Usenet protocol, regardless of what a very-rarely-used RFC
about X.400 gatewaying has to say about mail messages.
Should such a thing ever be offered by the protocol and supported by any
server, one of the first things that I'd add to my filters would be rules
to reject any message with multiple message IDs in Supersedes, at least
until we have some sort of authentication mechanism.
-- Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>