First problem: I don't think that can be done. I'd be a lot more convinced to consider arguments like this accompanied by completed samples in some real area of human endeavor of the "black and white laws" with accompanying "unit tests" (and, anyway, wouldn't we want automated user acceptance tests, not unit tests.)