> 1. Should we really the ask the question that how many of these test
> sequences are useful? what I mean to say is that all of these tests
> are excercising the system in all possible state. So if we ask the
> question as how many of these tests are useful, are we not
> restricting the number of states?

Depending on how you will use the model for test generation purposes
the number of test cases can not be so huge. For instance, I use UML
models to generate tests using some well-known criteria like
equivalence partitioning, boundary value analysis and some others. The
amount of test cases is equivalent to the number of test manually
generated by a human tester. So I do not have this problem using MBT.

[]'s

Pedro.