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

On Tue, Aug 03, 2021 at 12:41:45PM -0700, Dave Crocker wrote:

> On 8/3/2021 12:30 PM, Viktor Dukhovni wrote:
> > It makes the localpart content of the "Delivered-To" header a private
> > matter for the MTA that adds the header.  It need not identify the
> > mailbox to anything other than the MTA.
> > 
> > For purposes of loop detection the order in which "Delivered-To" headers
> > are added is not material, so requiring it to be prepended, as though a
> > trace header, may not comport with some existing implementations.  FWIW,
> > in Postfix it is in fact prepended.
> 
> 
> 1. How does that affect the technical specification?

    * The *requirement* to prepend (a la trace header) is not essential
      to the established use-case, so could be omitted.

    * The content of the localpart need not leak the raw address to
      which the message was delivered.  Various kinds of localpart
      obfuscation are allowed, and can be desirable.

> 2. When there is a sequence of Delivered-To header fields, for a mail 
> that transits one or more intermediaries before 'final' delivery, how 
> does this affect utility of the field?

The set of headers form a set, not an ordered list, a match indicates a
loop regardless of position in the list.

-- 
    Viktor.