The simple solution would be to have independent entities offer trust assertions about CAs and to allow users to consider multiple entities' views in their decision about whether to trust. It's surprising this doesn't exist yet when the attack vector is so clear.
replies(3):