Wondering whether http://tinyurl.com/2fa3kn should not better be two specs, one for the protocol level, one for the participant semantics