Re: [ietf-822] WSJ/gmail/ML, was a permission to... (on-topic)

[Alessandro Vesely <[e-mail]>]

On Tue 06/May/2014 03:30:30 +0200 Russ Allbery wrote:
> Miles Fidelman <[e-mail]> writes:
> 
>> I haven't actually dug into the details of how Outlook does things,
>> but... does not RFC5322's series of resent- headers start to provide a
>> direction for standardizing mailing list use of header fields?
> 
> I think these are used in the opposite direction of what you'd need to
> satisfy the current constraints.  The entity doing the resending goes into
> the Resent-* headers, and From is left unaltered, whereas to satisfy this
> signature scheme you would need to do the opposite.

I beg to differ.  To adjust the signature scheme so that it works in
the face of resending is plan A.  The From: field is set by the
author's MUA and checked by the MSA.[1]  Leaving it unaltered is a
privilege that resenders need to earn by enforcing MSA-equivalent
checks.  WSJ article sending is an example where From: ought to be
changed, while gmail and MLs can keep it unaltered.

It is a technical challenge to define authentication correctly, but we
should not modify the semantics in order to meet the constraints.
This problem is not specific of DKIM signatures.  S/MIME and OpenPGP
present it too; for example, Thunderbird fails to verify S/MIME-
signed mailing list messages[2].

Whitelisting by (sub)domain name can be done according to how well
they carry out the checks they're responsible for.  Maintaining
whitelists without relying on authentication --plan B-- will likely
require more human knowledge and personal judgment than with a working
signature scheme in place.

Ale

[1] http://tools.ietf.org/html/rfc6409#section-3.2
[2] https://bugzilla.mozilla.org/show_bug.cgi?id=885286#c4