That is well said, Bron. For the record, I have no objection to DKIM2, and I do hope it helps. I do have concerns that it is being oversold.
The difficult filtering problems involve absence of information. DKIM2 solves some problems when it is present and verified. What does an evaluator do when it is absent or fails to verify due to DNS timeout? SPF and DMARC attempt to solve that dilemma by distinguishing between hard fail and soft fail, but the distinction does not work. A domain owner cannot usefully advise a stranger he does not know, about how to handle an email message that he has not seen. My goal with ARC was to do SPF correctly: find the originating server organization and the originating SMTP MailFrom domain, then apply the SPF test to it. If the test fails, but the message stream is acceptable, then create a local policy that links the server organization to the MailFrom organization. Once that process correctly authorizes all of my wanted mail, then I can more aggressively quarantine messages that might be impersonations behind intermediaries. The analysis has proved difficult. It is difficult to determine the originating server organization reliably, and difficult to determine the original MailFrom organization reliably, due to inconsistent and missing information. I am currently verifying SPF-equivalance, at origination, for about 50% of my accepted indirect traffic. Getting the rest authenticated will require many more local policy entries. ARC helped this effort by providing data. The Source IP and Mail From information has been reliable, even when authentication test results were not. My reluctance to see ARC go away is because it makes data go away. DKIM2 does not expect to distinguish between originating server and outbound gateway, so the quest for authenticated originators will remain unsolved. Doug On Tue, May 19, 2026 at 8:39 PM Bron Gondwana <brong= [email protected]> wrote: > > > On Tue, May 19, 2026, at 19:28, Douglas Foster wrote: > > Bron, > I was speaking from the viewpoint of a system admin on the receiving side, > DomainB, who is putting controls on inbound forwarding. There are many > reasons why self-configured forwarding is problematic. Forwarded messages > that are judged unacceptable by the recipient domain will put the > forwarding domain or server at risk of being blacklisted. Since every > organization has different filtering rules, it is pretty much guaranteed > that some of the forwarded traffic will be judged unacceptable, and loss of > credentials during forwarding increases this probability. Malicious > forwarding configurations can be used to harass the forwarding target. > Mistaken configuration of forwarding can harass an unwitting victim. > Unauthorized forwarding from business organizations creates a significant > risk of information leakage in violation of privacy regulations. > > I was responding to a couple of posts that envisioned a world where > forwarding was arranged by mutual consent between domains and there users. > Doing so would help to mitigate those risks, but it also runs into > problems. > > > That's a lot of kind of general statements that I don't think illuminate > much, and it's really hard to design from them. > > Yes, you identify some considerations a system administrator has when > accepting forwarded messages. > > "Email is hard, let's go shopping" -- Barbie, circa 1995, probably. > > It's hard. I acknowledge that it's hard, and DKIM2 isn't going to a magic > bullet that makes everything easy. > > We're trying to build a system that makes some difficult problems more > tractable. I haven't seen anything in your responses so far which either > propose a better way to solve said problems, or show that DKIM2 will make > said problems harder than the currently are. Just that "they will remain > hard" > > I agree. They WILL remain hard. But at least DKIM2 will help the > receiving system look through the (user configured) forwarder, and more > reliably know whodunnit on the unacceptable forwarded traffic. > > Which means that a well behaved forwarder doing DKIM2 is *less* likely to > be blacklisted (sic) than under previous arrangements, because the > recipient can tell that they are innocent of the wrongdoing. That's not > nothing. Sure it's not perfect, particularly during the early stages, but > it's positive improvement over throwing up our hands and not trying. > > Bron. > > -- > Bron Gondwana, CEO, Fastmail Pty Ltd / Fastmail US LLC > [email protected] > > > _______________________________________________ > dmarc mailing list -- [email protected] > To unsubscribe send an email to [email protected] >
_______________________________________________ dmarc mailing list -- [email protected] To unsubscribe send an email to [email protected]
