On Mon 08/Jun/2026 23:00:55 +0200 Murray S. Kucherawy wrote:
On Thu, May 28, 2026 at 8:58 AM Alessandro Vesely <[email protected]> wrote:
### The Missing Authorization Layer
ARC, much like DMARC itself, ultimately leaves delivery decisions to
local receiver policy. A receiver evaluating a DMARC failure may decide
to override the domain owner's published policy when a valid ARC chain
is present. In practice, this override depends on the degree of trust
the receiver places in the ARC sealer.
From the perspective of intermediaries such as mailing lists and
forwarders, this behavior is fundamentally unpredictable. A forwarder
has no practical way to determine whether a receiver trusts a given ARC
sealer, or under which conditions that trust applies. As a consequence,
mailing lists largely abandoned reliance on downstream ARC evaluation
and instead chose to assume responsibility for the message in the DMARC
sense, typically by rewriting the From: header field.
I've read this over a couple of times, and I'm not convinced that the
proposed text adds something that's sorely missing.
My concern is that, without identifying the missing component, the reader is
left with the impression that ARC's shortcomings were primarily in the
mechanism itself rather than in the trust model around it. The draft already
attempts to extract lessons from ARC for future work. My proposed section is
an attempt to articulate one such lesson.
If we feel like we really need to expand on what might have made ARC
better, IMHO these two paragraphs say enough.
The remainder of the proposed section is an attempt to explain that observation
and to argue that whatever additional mechanism would have been needed to make
indirect flows predictable could reasonably be characterized as an
authorization layer.
I'm not totally comfortable with the last sentence in the second paragraph
though, because I don't know that list developers or at-scale operators
even gave ARC anything beyond a cursory review; the From: field rewriting
may have been the path they chose to follow independently.
I agree with your concern. "Largely abandoned reliance" implies a level of
evaluation that often did not happen. What I was trying to capture is that
intermediaries could not safely rely on downstream ARC evaluation, regardless
of whether they explicitly considered ARC or simply adopted the more
operationally predictable approach.
Perhaps this is better:
As a consequence,
mailing lists could not rely on downstream ARC evaluation and instead
chose to assume ownership of the message in the DMARC sense, typically
by rewriting the From: header field.
Best
Ale
--
_______________________________________________
dmarc mailing list -- [email protected]
To unsubscribe send an email to [email protected]