Re: homework, not an experiment, draft-crocker-email-deliveredto

On Fri, Aug 06, 2021 at 11:56:50AM -0700, Dave Crocker wrote:

> > If an MUA wants to learn the envelope recipient address to which the
> > message was sent (by the immediate upstream MUA or relay) then Delivered-To
> > is not the right header for that.  That's why Exim has Envelope-To, while
> > Postfix has "X-Original-To:", both of which are intended for consumption by
> > tools like fetchmail.
> 
> So you are saying that it is not possible for one Delivered-To to show 
> the actual address it has been delivered to, followed by another 
> Delivered-To that shows a local transformation to an internal delivery 
> string?

Well, since "Delivered-To:" is for MTA-private use (loop detection) and
(by default) records the internal envelope recipient passed to the LDA,
rather than the input envelope recipient from the sending agent, it was
deemed appropriate to conflate these into a single header.

> Or are you saying that Delivered-To has nothing to do with delivery?

I am saying that "Delivered-To" is used for detecting loops in case the
LDA reinjects the message back into the mail handling stream and as a
result it unfortunately arrives to again be delivered to the same
recipient.  The the simplest way to fingerprint the prior delivery was
to record the envelope recipient as conveyed to the LDA, but other
fingerprints are possible, and may be used in the future.

If the goal of the experiment is to repurpose "Delivered-To" as a
standard replacement for "Envelope-To", "X-Original-To", and have
that header now record the input (as received) envelope recipient,
rather than the current output (as passed to the LDA) envelope
recipient, that should be more explicit in the draft, and there
should be a compelling case for Qmail, Exim, Postfix, ... to
make that change.

-- 
    Viktor.