Re: sender rewriting scheme
"Alan DeKok" <[email protected]> Wed, 08 Oct 2003 13:03:13 -0400
| Newsgroups | gmane.ietf.asrg.rmx |
|---|---|
| Message-ID | <[email protected]> |
"Gordon Fecyk - Home" <[email protected]> wrote: > > b) MAIL FROM contains the username at that domain who consented > > to forwarding the message. > > b) suggests that the sender envelope changed somewhere in the .forward > process. I don't think this happens with MTAs that support .forward files. > If it did, then .forward wouldn't be an issue. The issue is that I *know* I have no .forward files on machines in China. e.g. I receive a message with a EHLO for a valid domain, and that domain says "yes, that IP is allowed to send email as me." If the MAIL FROM then says "[email protected]", I *know* it's spam. If I had a .forward file in that domain in China, then I would have to accept the message, as probably being forwarded on my behalf. This means that ANY use of MAIL FROM means that MTA's will need to have white-lists which are applied *before* checking the MAIL FROM, but after checking the EHLO. e.g. "bob has a .forward file in china.com, so don't check MAIL FROM for messages to bob with EHLO china.com" Yuck. The use of DNS-based filtering on EHLO/HELO has many fewer problems and issues than that for MAIL FROM. > > EHLO [127.0.0.1] > > > > is conformant to RFC 2821. Yet it's behaviour is obviously > > nonsense. ... > If some MTA did that and it wasn't actually trying to connect from localhost > to localhost, DMP and SPF don't care since they look at MAIL FROM first. [1] My contention is that the entire envelope has to be checked as a whole, so we can't have the concept of order in the checks we do. We probably need to come up with a 3-tuple describing EHLO, MAIL FROM, RCPT TO, and go through the variations of 2 remote domains, 1 local domain, and various IP's. e.g. for (EHLO, MAIL FROM, RCPT TO), (ip, ip, local) = probably a spammer (domain1, domain2, domain3) = relay attempt (domain1, domain2, local) = message is .forward'd (domain1, domain1, local) = message from user at domain1 ... etc. Such a list would help us describe which solution applies to what situation. I'll try to type up something today. Alan DeKok.