This paper provides a discussion of structural test coverage metrics, looking at the practicality of their use for real software developments. It concludes that some metrics are unsuitable for real use, and recommends suitable combinations of str