Hacker News Reader

The comments to that story are pretty good.

An interesting question is: why does TDD fail in such experiments (it does so unexpectedly consistently), even when many developers feel it has benefits when they practice it?

There is no silver bullet, so there must be circumstances in which TDD does not work. And conversely, the central question is: under what circumstances does TDD work? What are the preconditions?

View on HN

The answer is likely in the studies, which have declared TDD efficient practice, and in the works that specified TDD approach. These works and studies tried to solve specific problems and I'm not sure they followed really scientific process before declaring that TDD is a solution, not a by-product of some solution that passed unnoticed during the research (e.g. education of developers on software architecture).

My gut feeling is that TDD is just a way to encourage me to actually write tests. I suspect I'd have just as many tests if I had someone prod me every 20 minutes to write a test case or two. Having one or two tests also creates a bit of a wedge to make me write more: once there's some code that can be neglected I can feel guilty when ignoring it.

Reply on news.ycombinator.com