Hi Craig... what you're worried about in a "new protocol" is exactly the
problem of the proposed "tweaked protocol". I don't see much difference.
The cost is the same, but the gains are vastly different. Furthermore,
most of your scenario disappears if there is an a-priori way to tell if
the remote is supposed to handle "new protocol" (and that you can therefore
wait for the service to be reestablished). This could be by fiat, using
magic cookie information, a ping of some kind, or whatever. Not a big problem,
as long as it is cheap to do offline.
Mumble... if I were a dictator (and the IAB is...), I'd just decree
that henceforth all 821 and 822 implementations MUST be 8-bit transparent.
In about 2 years (2 months in Europe), the vast majority of installed systems
will be compliant, and the only effort is that a bunch of people need to rip
out the code that prohibits 8th-bit-set characters. Machines that don't
natively handle 8-bit characters get to do the translation themselves, at the
last possible moment. That should be an incentive in some direction.
rayan