The thing I don't like about TDD is I never know if I'm doing it correctly. There's always that No True Scotsman fallacy^1 situation. "Oh, you tried TDD and it didn't work out for you? If you were doing real TDD it would have worked out." And I say this as a person who has watched three of Uncle Bob's videos on TDD that you have to pay for: http://cleancoders.com/episode/clean-code-episode-6-p1/show http://cleancoders.com/episode/clean-code-episode-6-p2/show http://cleancoders.com/episode/clean-code-episode-19-p1/show
Sorry for hijacking the topic. I've been trying TDD for years, and I never know if I'm doing it correctly. I saw this line and it made me question it yet again.
^1: http://rationalwiki.org/wiki/No_True_Scotsman
EDIT: Sorry, I commented as I read. Looks like you're going to explain it in the article.
Pairing in person with someone who's done it for a long time can help a lot, but even then you're getting someone's interpretation. If you worked with the DHH of three years ago, you'd get a very different interpretation (slower cycles, integration tests, no isolation) than if you worked with me (faster cycles, integration/unit mix, biased toward isolated unit tests). We were both doing TDD, though! (Given that he's said that he used to do TDD, I assume that he was following the core red/green/refactor loop.)
Fakes get you around the question of database integration, but not the question of integration in general. You'd have to create a fake of every class in the system for that, in which case you probably just re-invented mocks and/or stubs.
I probably could've more explicitly called out the fact that DHH is obsessing over database integration when it's just a special case of what isolated testers are actually worried about, which is integration in general. It was already a long post, though.