Thanks for these comments.
When a set of mail applications are bundled under a single umbrella, then the operator can do more - and the Transformation using an existing Alias is possible.
But, if a mail user is using discrete components, then can you reasonably expect them to interact as we suggest transformation?
Are we back to square 1' - having eliminated an algorithmic transformation of a non-ascii mailbox name?
FYI, the IETF's EAI group went through all of this stuff in excruciating detail. The original experimental version had a bunch of downgrade features, nearly all of which went away in the later proposed standard because they just didn't work. I think it's safe to say that the answer to any question of the form "how about if you downgrade using approach X" will be that "X turns out not to work, you'd better use the effort to make EAI work." The only downgrade features left are ones to let old clients fetch EAI messages from POP and IMAP servers, and they don't try to make anything relplyable, just smash the headers so they're valid under RFC 5322. Regards, John Levine, johnl@iecc.com, Primary Perpetrator of "The Internet for Dummies", Please consider the environment before reading this e-mail. https://jl.ly