Read "NullAway: Practical Type-Based Null Safety for Java" and
"Rapid: Checking API Usage for the Cloud in the Cloud", and then
answer the following reading questions:

1. NullAway aims for no false negatives in practice. How does this goal
differ from the standard formulation of soundness that we've used when discussing
other kinds of tools (such as abstract interpretations)? Does one of these
definitions imply the other?

2. Explain what a "pure" method is? Why does NullAway assume that all methods
are pure? If there is an impure method, what impact might that have on NullAway's
results?

3. What choice made by the NullAway authors
causes the 64% of NPEs in section 8.3 of the NullAway paper to not be caught by
NullAway? Could they have designed the tool differently and caught these errors?
If so, how? If not, why not?

4. What is a "malpractice" state in the Rapid paper? Does the notion of a malpractice
state relate to any of the design decisions taken by the NullAway authors?

5. Is specification fingerprinting sound? Why or why not?

6. What was the most confusing part of the reading? Is there anything you still don't understand?

7. How much time did you spend on this assignment (both the reading and answering the questions, combined)?