On Tue 28/Jul/2026 10:43:29 +0200 Laura Atkins wrote:
On 27 Jul 2026, at 11:23, Alessandro Vesely <[email protected]> wrote:
On Wed 22/Jul/2026 12:29:50 +0200 John Levine wrote:
It appears that Laura Atkins <[email protected]> said:
The other data piece I think will make it clear about the adoption of ARC is
related to mailing lists. My impress is that most mailing list manager software
chose to rewrite the from address rather than sign with arc.
I wrote the ARC support for the widely used Sympa list manager, and what I've
seen is that lists put on the ARC headers but they rewrite the address because
ARC doesn't work.
Rewriting From: and adding ARC headers are two quite different things. They
are not alternatives to one another. Rewriting From: is a hack used to avoid
DMARC policy failures. Adding ARC headers is an attempt to participate in a
collaborative mechanism. There is no choosing one rather than the other.
A mailing list cannot know whether a recipient's MX will trust its ARC seal, or
what it will do with the information conveyed by ARC. It therefore cannot
safely refrain from rewriting From: on the assumption that ARC will preserve
delivery.
Thank you for so clearly restating my point.
ARC was not sufficient to preserve delivery and so mailing lists chose to
rewrite from addresses to avoid the collateral damage of bouncing subscribers
who rejected mail that failed DMARC.
The immediate cause of that choice was not that ARC was insufficient by itself,
but that DMARC deployment made delivery failures an immediate operational concern.
That, I think, is the missing part of the ARC experiment. ARC provided a way to convey
authenticated information across intermediaries, but did not provide the authorization
needed for an intermediary to rely on receivers accepting its transformations. In that
sense, saying simply that "ARC doesn't work" misses an important point: the
mechanism needed to make it work for the mailing-list problem was never completed.
Work wasn’t completed in more than a decade AND there is a widespread hack to
work around the failure to complete the work. That says that the ARC experiment
failed and we should use Trent’s draft as a starting point to just say that and
move on.
I think we're still talking about different questions.
The fact that mailing lists continued to rewrite From: is not in dispute. The
question is why.
ARC did not give a mailing list any basis for assuming that a receiver would
accept its modifications. That is an authorization issue, not an
authentication issue. ARC authenticated what happened to the message, but it
did not provide a way for the intermediary to know that its actions would be
accepted by receivers.
Whether that missing authorization piece should ever be specified, or whether
From: rewriting should simply remain the de facto standard, is a separate
question. My point is only that, if we're documenting why ARC did not solve
the mailing-list problem, I think that distinction is worth recording.
Best
Ale
--
_______________________________________________
dmarc mailing list -- [email protected]
To unsubscribe send an email to [email protected]