> So if both sender and recipient are using this, they'll end up demanding increasing classifications from each other in an infinite loop?
Presumably the system would only kick in for threads of discussion initiated by others and not yourself. So if you send an email, any replies to it would behave as normal.
That requires "standards" compliant behaviour which won't happen. Similar issues happened back in the days when people used to respond with "click this link to verify" as ill thought out anti-spam measures. The inboxpro proposal is also vulnerable to backscatter, joe jobs etc.
Presumably the system would only kick in for threads of discussion initiated by others and not yourself. So if you send an email, any replies to it would behave as normal.