Now, while I love the name of the klensin-has-flipped-out header, and I
hope that John reaches some kind of immortality some day, I really hope
that he has indeed paranoiaed-out, and that he shouldn't worry so much
about what 8-to-7 converters might do. I think it's perfectly
reasonable for the message source to make a suggestion for a 7-bit
conversion algorithm, but to bounce if the suggestion is absent?
Old story, no need to restart it.
But I think that Greg's suggestions, and several of their
predecessors, imply that they must *do something*, and do something
deterministic. So "if you can't figure it out, bounce" really isn't
new.
And I had assumed that "all converters must know all encodings" was
implicit in where we are headed, just as "all converters must know how
to parse messages" is. The latter is the stronger requirement; "all
encodings" is a total of two at the moment.