Dave Crocker <dcrocker(_at_)brandenburg(_dot_)com> wrote:
Who has authority to set the mailfrom?
The administrative domain which "originates" a message. I'd say
then that an end-user "submits" mail to that domain for delivery to a
destination.
If more than one entity has the authority, what is the relationship
among them?
Only one entity: the administrative domain which will accept
bounces.
Note that mailing lists can be viewed as an end-user using SMTP to
deliver a message, which is then submitted to the list exploder. The
administrative domain for the list then accepts responsibility for the
message, the bounces, and thus sets MAIL FROM.
If we validate that the field is authentic, what good is that? What will
be better? What will not be changed?
We will know that someone has accepted responsibility for that
message, and that a bounce path exists. To be pedantic, any LMAP
implementation should also check the MX records for the bounce path.
If here are no MX records, then the bounce cannot be delivered, and he
message should not be accepted.
What will be better is that the restrictions of RFC 2821 will be
better enforced. An MTA should either accept responsibility for
delivering the message, or responsibility for bouncing it. If it
can't bounce the message, it has no failure path if it later has an
error (spam or not).
To put it another way, RFC 2821 requires MTA's to deliver or bounce
the message. However, it provides insufficient methods for an MTA to
determine if it is possible to bounce the message. Additional
verification can then be viewed as fixing design flaws in the original
proposal: There is a valid use-case of SMTP in which an MTA must
contradict the requirements of SMTP, and fail to deliver or bounce a
message, after it has accepted responsibility for it.
Alan DeKok.