>> This seems odd.
> I fail to see anything odd about it.
Don't you think it's odd to explictly allow labeling non-822 addresses
No. It's a question of priorities. THey're supposed to be valid RFC 822
addresses, but having an accurate indication of the original address - legal or
not - trumps the desire for syntactic conformity, especially when enforcing
that conformity has the very real possibility of breaking the key function
this field provides.
I note in passing that we do this sort of thing with names all the time. Just
as one exmaple, a message/rfc822 MIME object doesn't have to conform to all the
rules RFC 822 imposes.
I mean, I see the argument for allowing anything, but to then label that
rfc822 seems very odd.
The name provides an indication of what it should look like. I see that as
having value in and of itself.