Rethinking BGP path validation (part 2)