[Raw Msg Headers][Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Zmailer SMTP: Anal retentive or correct?

At 15:23 7/24/96, Scott Ballantyne wrote:

>I would argue that it's a bug in the sending ZMAILER - rfc-821 says
>that the argument to FROM: should be a reverse path suitable for
>mailing an error message to.
>So the sending zmailer should turn the simple FROM:<user@uucp> into a
>source route, and in fact rfc-821 says that:

See RFC 1123 (Host Requirements, Part II). The route-addr syntax spec'd in
RFC 821/822 for the return-path in SMTP has been explicitly deprecated, for
several years. No one is supposed to generate it any more, and if you do,
by chance, get one, you're perfectly allowed (nay, encouraged) to short-cut
it straight to the user@host part.

The implication of this is that *all* addresses in headers and in the
envelope of Internet mail MUST have Fully Qualified Domain Names (FQDNs) on
them, and it is the responsibility of the host (gateway) to modify a letter
from any other E-mail system (UUCP, QuickMail, MHS, whatever) into Internet
canonical form. Server SMTP implementations are certainly within their
rights to check and reject any E-mail which does not conform to the
Internet standards, particularly when, as in the case being discussed, such
E-mail cannot be replied to by the unfortunate recipients (after all,
they're only required to follow the Internet E-mail standards, not worry
about what UUCP hosts someone else is connected to).

There are several ways to transform UUCP E-mail addresses to the Internet
(and vice versa) successfully:

        host!user -> host!user@gateway.foo.com
        host!user -> user%host.uucp@gateway.foo.edu
        host!user -> user@host.org (provided an MX to gateway exists)

All of these are completely legal 822 addresses; if you're not the host on
the right hand side of the "@", you have no business attempting to parse
the local-part. If you *are* the host on the right hand side (or an MX for
them), one hopes that you will do the right thing with whatever local-part
you emit when it comes back to you in a reply.

The real rule is simple: the gateway host must transform the letter into
Internet canonical form, and be prepared to do the right thing with a reply
to such a transformed letter when it gets dropped back into gateway's lap
(i.e. the transformation must be reversible, preferably with no loss of
information). Failure to do so is the responsibility of the postmaster of
the gateway host.

        been there, done that, have the scars to prove it,

        Erik E. Fair <fair@clock.org>