Stateless Tests in a Stateful Protocol

January 25th, 2010

Lets suppose you want to test the correctness of a protocol. I know I do, all the time; get me that input fuzzer. A typical way to start might be a minor variation on how I learned to test my code in CSE 142 and hadn’t really thought too much about since: plug in some “representative” values and see if it does the right thing. But let’s suppose someone came over with a pile of documentation and said “Here. Some genius in the back room made this insanely complex server that apparently speaks this protocol, and somehow managed to do it without so much as a single unit test. Can you test it? By the way, we want some guarantees about security.”