John Leslie wrote:
Dave Crocker (and others) set out this idea in Bounce Address
Tag Verification: http://mipassoc.org/batv/
Yes, I know this. I try to explain why the term "bounce address"
in the context of Dave's I-D is IMO an excessively bad choice:
1 - You've already noted that RFC 2821 still uses "bounce" as a
synonym for "reject". RFC 2821bis will fix this, and UTF8SMTP
avoids this confusion. But we're not there yet, Dave's I-D
has to make sense for readers only familiar with RFC 2821.
2 - BATV by design enforces a sound concept of a "BA". Arguably
SES, SRS, and redistribution (incl. mailing lists) are also
built on a sound idea idea of BA.
Besides the BA in BATV is actually a shorthand for "crypto
timestamp encoded envelope sender address allowing to
reconstruct the original envelope sender mailbox in the ADMD
of the originator" (for Dave's definition of ADMD). That's
a mouthful, and BA is a nice shorthand for this concept.
OTOH Dave's BA cannot guarantee that it's in the ADMD of the
originator, it can be in an unrelated ADMD of a 3rd party
without prior consent of its administrator. For various
reasons that could be very wrong, it also doesn't reflect
the original idea of a hop-by-hop return path.
3 - Last but not least Dave claims that the term "return path"
was always wrong, it should have been BA. IMO that's not
the case, quite the contrary, 821 SMTP was well designed.
It was the removal of reverse routes while keeping 1123
5.3.6(a) forwarding that caused in essence most problems
with mail today.
IOW, the established term "envelope sender address" is far better
than "bounce address", the former is correct, the latter should
be limited to contexts like BATV.