There is a number of points to note about the process that has been sketched here for the part I requirement (Functionality requirement -- part I). First, it is designed to provide a way of combining and comparing information from various parts of the task model in order to express the requirements in more detail. Secondly, it should be made clear that not all parts of the model need be individually present; many values can be derived from defaults, aggregate scores and so on. It is also the case that in practice we will often not be able to make all the distinctions present in the model; the model indicates what ought to be considered, but practical difficulties may interfere. The problem of ranking advice according to end-user understanding, when advice from grammar checkers comes in so many different and non-comparable forms, is an example. However it is felt that stating the requirements in this way helps to clarify the issues, and, for example, allows us to motivate differences between the evaluation methods possible for grammar and spelling checkers.