Log file format is in any event a locale-specific issue, so it's not clear
that it would be appropriate for an MTA to log an unencoded UTF-8 address to a
log file on a system that used some other charset in its locale.
First, it's not a system that uses a locale, it's just a certain user/application/whatever.
In addition to that, assuming addresses in many different (languages and) scripts, you better use UTF-8 or another encoding that is able to encode all of Unicode/ISO 10646.
> Essentially this proposal appear to add a i18n-layer for e-mail, a
> "presentation" layer if you wish, on top of RFC 2822 and SMTP. While
> this may be preferable for ASCII people, having i18n as an "add-on"
> appear rather fragile to me. But that's only my initial reaction.
Even if we were designing a new mail system from scratch to handle these
requirements, I think we would need a lookup across the net at the time
that the message is composed (or submitted) in order to handle the case where
at least one recipient can't transcribe the sender's "native" address.
As I tried to explain in another mail, I'm not really clear on this. The sender should not send an address that the receiver doesn't know how to read.