self destructing messages will always be trust based. it's fine for certain threat models but not others