On Wed, 30 Jul 1997, Bart Schaefer wrote:
Put another way, is there ever a case where a sending agent needs to start
from a primary address, compute a subaddress by some means other than user
input, and direct the message to the concatenation of the two?
A lot of sites have a requirement that the from address not be editable in
their MUAs. Such sites do this because they don't want realistic-looking
spoofing to be easy. I can't think of anything the IETF could do to
change this. We can jump up and down and say "you must be able to edit
the from address" all we want and it simply won't happen.
So I'm looking for the ability to enter just the subaddress for the From
header and envelope from without any requirement that the rest of the from
address be editable. I bet MUA authors would go along with that (or a
special list-subscription command which added a subaddress) and it
wouldn't violate site policy for most sites which don't want the from
address editable.
At this point I don't find the arguments for standardizing to be strong
enough, though an informational document might be nice. I'm willing to be
convinced otherwise, though, because I do get mail via several domains and
do feel the inconvenience of inconsistent or absent subaddressing schemes.
There are two cases where MUAs need to cooperate with subaddresses:
1) MUAs which don't allow from to be edited otherwise.
2) MUAs which validate local addresses against a white pages or authdb
service.
There is one case where a list server needs to cooperate with
subaddresses:
1) List server restricts posting to subscribers.
There are several cases where a final delivery agent needs to cooperate
with subaddresses:
1) alias or forwarding address (especially for centralized mail hubs which
route delivery to other nodes)
2) validation of local addresses against locally deliverable address list
3) Any final delivery filtering/filing rules.
I believe this demonstrates the need for standards-level interoperability
between the three components.