I stopped reading when the author basically says that what a lot of people consider to be unit-testing is not important, which is fine, but doesn't counter the other side's argument. I believe there is a benefit for making mock objects to improve speed. The author says it's not important, doesn't explain why other than the saying that he doesn't care.

That's about when I stopped reading.

TestNG creator Cédric Beust contradicts himself on one of his main points. First he says "Unit testing is an important part of testing..." but further down claims that "It's not." The point he makes here is that JUnit tests should be allowed to make database calls and mock objects should be avoided.

There is no argument here. JUnit does not preclude you from making database calls. In fact you should have both tests with database calls (integration testing) and tests using mock objects (logical unit testing). Each one of these serves a very different purpose. Vincent Massol covers this in good detail in his book "JUnit in Action". As for mock objects, beyond merely stubbing out expensive calls they can provide further benefits such testing interractions between the object under test and the classes it depends on.

There are also a number inaccurate statements throughout this interview that should be pointed out for readers. For example Cédric Beust implies there is no way to initialize expensive objects in JUnit once around test classes or suites. In fact this can be done with the TestSetup class in the JUnit framework.

As for not being able to swap tests in and out without recompiling that doesn't bother me. In the agile way of thinking all tests should run as often as possible, which is another reason to avoid expensive calls in logical unit testing.