ietf-822
[Top] [All Lists]

Re: Understanding response protocols

2004-09-11 20:32:58

Bruce Lilly writes:
Reply-To: ietf-822(_at_)imc(_dot_)org / From: djb(_at_)cr(_dot_)yp(_dot_)to 
does work

We could certainly use the name Reply-To instead of Mail-Followup-To in
a completely new mail system.

The problem is that we're not designing a completely new mail system.
Your proposed protocol, sending followups to Reply-To and replies to
From, has horrible effects when it interacts with today's software.

Specifically: A huge number of today's messages have a suboptimal reply
address in From, the desired reply address in Reply-To, and the desired
followup addresses in To etc. What happens when these messages are
interpreted according to your protocol? Answer: not only do replies go
to a suboptimal address, but also followups are completely misdirected.

``But the desired reply address could go into From, leaving Reply-To
empty!'' you scream. Yes, it _could_ go into From, but in _reality_ a
huge number of today's messages have it in Reply-To. Moving it means
changing and redeploying a huge number of programs, including homegrown
mail scripts that haven't been touched for years. Only _after_ that's
done would it be safe to start using the name Reply-To for something
else---in particular, for MUAs to switch from To+Cc+(Reply-To||From) to
Reply-To||(To+Cc+From) for followups.

I'm not saying that your protocol is inherently broken. I'm just saying
that it's disastrously incompatible with current practice.

For comparison: Mail-Followup-To avoids all the costs of that first
transition step. It's safe for MUAs to switch to Mail-Followup-To||...
for followups; many MUAs have done this already, saving time for their
users. The use of a _new name_ makes the Mail-Followup-To protocol much
less expensive to deploy than the protocol you're advocating.

---D. J. Bernstein, Associate Professor, Department of Mathematics,
Statistics, and Computer Science, University of Illinois at Chicago

P.S. I pointed all this out, although not in as much detail, in my first
``Understanding response protocols'' message. This discussion would make
more progress if you started paying attention.