Welcome to the third episode of Is TDD Dead.

I'm Kent Beck from Facebook, and I'm here with David Hannah Meyer Hansen and Martin Fowler.

And we're going to follow up on episode two with a conversation today about the trade-offs inherent in testing.

I think of most interesting questions in software development

is involving some kind of trade-off.

And I do this with my hands, which I understand is remind some people

of some sort of bizarre form of modern dance, which is your problem, not mine.

But what I mean to convey with that is that there's some set of costs in one side and another set of costs on another side,

and there's some minimum between those two curves.

And depending on the way that the two cost curves are shaped,

will change how exactly you decide to go on some continuum.

Almost no questions in computer science can be answered A or B.

The answer always begins with it depends and what does it depend on is the interesting part.

Now in some ideal world we would have instant infallible feedback about our programming decisions.

So every keystroke that I make, if the code is ready to deploy,

it would just instantly deploy within milliseconds.

So if I just happened to create some software that works, then boom, it's instantly

out to everybody. And that kind of feedback is impossible at the moment. By calling my

methodology extreme programming, I learned you can't ever say never, because what seemed extreme at the time turned out not to be.

But that's the ideal world, that you get this instant feedback about whether your software is ready to go or not.

And the reality is that we have to back a ways off of that.

The question I think is how far off of that do we need to back?

...