Re: rough sketch of a potential solution

> Another issue, if I understand the proposals right, is that the
> actual addressing is always done on the pure ascii-only address
> in the original From:,... fields.

Not exactly, because the ascii-only address (which  could either be a 
pure-ASCII alternative or an IMAA-encoded address with characters not 
limited to the ASCII repertoire) is automatically derived from whatever 
address the sender typed in.

> So in one way or another, the
> alternative addresses are just comments or decorations, and they
> are just sent so that we don't have the recipient to have to look
> them up, so that we can avoid privacy issues. That somehow doesn't
> look like much different from the current 'comments' facility.

One difference is that any of the alternative addresses can be used to 
look up the rest of the alternative addresses as well as the designated 
"fallback" address.  So none of the addresses are really comments.  
This is more like multipart/alternative for addresses than a comment 
facility.

Other reasons (besides privacy) for not having the recipient's MUA look 
up alternatives are efficiency and scalability - one lookup per message 
sent is a lot cheaper than one lookup per message read.  Also, 
deferring the lookup until the time that the message was read (perhaps 
years after composition) would in some sense change the meaning of the 
message for a lookup that yielded different results than it would have 
produced at the time the message was composed.

The other reason for doing the lookup at composition time is that it 
can then become part of the user interface - potentially allowing 
correction of minor spelling errors or substitution of alternate forms 
of an address or perhaps even sender selection between multiple 
recipients with similar names.