Wietse:
What I describe would be a best practice application of DKIM
mechanisms that already exist.
Mail is signed as if there are N+1 instances of each header that
is covered by the DKIM signature. The verifier will then fail if
any such header is added after-the-fact.
With this, there is no need to rely on enforcement mechanisms
outside DKIM, such as the correct implementation of RFC 5322.
Murray S. Kucherawy:
I would suggest constraining that to include only those fields
that are 0-or-1 in RFC5322 Section 3.6. For example, doing this
with Received: is begging for signature invalidation on otherwise
unaltered messages.
I see your point, but there are more "sensitive" headers than the
0-or-1 headers in RFC 5322 (IIRC, the N+1 signing method was
introduced to protect MIME headers).
I suppose that the guidelines for best practice application of DKIM
could recommend what headers to sign with the N+1 signing method.
These guidelines can be updated as RFC 5322 evolves, and as standards
that extend RFC 5322 introduce new "sensitive" headers.
Wietse
_______________________________________________
NOTE WELL: This list operates according to
http://mipassoc.org/dkim/ietf-list-rules.html