On Tue, Feb 22, 2005 at 02:21:06PM -0700, David MacQuigg wrote:
Why should a sending host also be a receiving host ?
Good point. The bounce should actually go to "postmaster@<sender's
domain>" where the sender's domain has been authenticated. Note: The
sender might actually be another forwarder, because we can't trust any
headers prior to that point, even if they say Authenticated.
Then why make it this difficult?
What do you think about this: http://www.spf.idimo.com/fix-2.php
At this point, I'm more focused on fundamental requirements than a
particular implementation. Get the requirements right, and we can avoid
lots of misunderstanding when we get to implementation. Is there anything
about the requirements we should simplify? Anything we should add?
That small piece of email which was posted to this website is
not ment to be read as a full implementation proposal.
It is ment to be an example, outlined by a techie for techies, of
a fundamental other approach to forwarding than SES, SRS and whatever.
Requirements: Mail needs to work.
Solution: Use something that works. This technique is in use today,
it just needs automation on both ends to make it scalable.
When one forwards mail to spamcop, one uses this technique.
If you read it carefully, you'll see it is flexible enough to
be implemented in various setups, it does not suffer from changing
headers (because the mail is payload, including its headers) and
it solves the bounce problem.
I'm not sure why it is being ignored. So far no constructive
comments have been made on this proposal yet everybody seems
to continue complaining about $foo whereas my proposal does not
suffer from $foo (or at least I don't think so).
Whatever solution we'll come up with, things have to change.
I think that using existing infrastructure and techniques
such as procmail and mime-encoded attachments are easy enough
to implement and with the least amount of hassle. Submitting
mail to spamcop works, usually without problems, and my proposal
is to use a similar technique.
Alex