Re: [EAI] [eai] #11: Capture the scenario of "SHOULD not use UTF-8 in Message-ID" for future advice draft

[[e-mail]]

> > Gateway operators are used to the idea that they have to do
> > something with Message-IDs, and while UTF-8 makes this job more
> > interesting it won't be completely different.

> Practical Mail->News gateways have to do all sorts of gross stuff due
> to news systems enforcing syntax rules more strictly than mail
> systems.  Munging message IDs would not be any worse than what they do
> now.

> Although I understand the rationale for wanting ASCII message IDs in
> mail to make non-EAI replies easier, I'm having trouble imagining
> plausible scenarios in which it would make a difference.

> Scenario 1: A sends B an EAI message.  B replies, and somehow manages
> to leave out all the elements that made A's message need EAI.  This
> suggests that A and B both have ASCII addresses, A's message was cc'ed
> to EAI recipient C who B leaves out of the reply, or something like
> that.  Yes, it could happen, but it doesn't impress me as likely
> enough to be a design point.

I somewhat agree and somewhat disagree. Replies to subsets of recipients are
very common. And my suspicion is that initially at least EAI usage will mostly
be in the extended address area. So I think sort of thing will come up a fair
bit.

OTOH, since the original message sent to these ASCII address recipients was EAI
and for good reason, what's the real value in being able to send a non-EAI
reply? For this to be useful there has to be some sort of routing difference
between B and A as senders involving non-EAI-capable systems. This, I think is
what makes this case obscure and not really worth worrying about.

> Scenario 2: A sends B an EAI message, something in the path uses local
> knowledge to downgrade the message to ASCII.  But if it's going to
> rewrite all the other headers, it can rewrite the message-ID, too.

Maybe, maybe not. It depends on whether or not it's an algorithmic rewrite or
not. Having to keep a mapping database is expensive.

> So I'm OK with it being a MAY, but it doesn't seem worth much effort.
> It's also one of those things where no matter what we say, MUAs and
> MTAs will make them UTF-8 anyway.

Agreed on both points.

				Ned