OneSuitePractitionersReport

The report came out of several (rather heated) discussions we had back in the autumn. You see, Geoff and his team are doing full XP (and I'm jealous) with the exception of UnitTesting, while I am doing pretty much none of XP except UnitTesting. The paper explains why in the end I admitted defeat, and was forced to contemplate that XP might possibly be done without unit testing. The big question was, would anyone else be convinced?

We submitted the report to the Xp2002 conference, and also to the WTiXP? workshop, which was billed as being "lively and entertaining" with "discussion and interaction". They even asked for papers that would "provoke discussion by presenting a controversial point of view". Great, I thought. Someone else will get a chance to explain to Geoff that his team is doing it all wrong.

The workshop was on the first day of the conference, before most people had arrived, and initially seemed to be living up to expectations. We had a good discussion going in the morning, but then broke into groups in order to search for "testing patterns", ie things we were all doing in our testing that was useful to be able to talk about.

In my group, we had a ball. We found we had loads of interesting testing patterns to talk about, and proudly wrote them up on large sheets of paper. Geoff's group, however, didn't seem to get along quite so well. They all seemed to have totally different backgrounds, and had trouble agreeing about anything to do with testing. However, instead of giving up and looking for "open issues for future work" or something, they pressed on, and poor Geoff never got to talk about his contravertial XP team. The rest of us were left wondering what had gone on, when they came and presented their "patterns" to the group. "Modularize your code - it makes testing easier". Yes. And?

However, all was not lost, we still had the poster presentation to go - the ideal chance for some lively debate, we thought. As it turned out, the posters were put up on boards in the sunshine by the pool. People were milling around, drinking coffee and feeling lethargic after the previous night's heady entertainments. Several people did stop to look at our poster, but the feedback we got was fairly calm. A few people took us up on points that we had omitted to write about in the report, but the general reaction was just puzzled. Wasn't anyone going to take us on and really debate the matter?

Just as we were about to pack up and go home, KentBeck happened to be wandering past - here was our chance! I grabbed him and proceeded to waffle excitedly about what Geoff's team was up to. He took it very well, considering, and even said that WardCunningham had said something similar, (about acceptance tests being more important than unit tests). What really impressed me, though, was that as we talked through what was going on, he very quickly spotted a weakness in Geoff's approach!

Geoff's team has loads of acceptance tests, and they largely take the place of unit tests, because they run fast enough to support refactoring. However, they havn't found a way of specifying the tests before they implement a feature. It hadn't occurred to them to try in fact. Kent said he thought a large part of the strength of XP testing comes from the fact that the tests come first - could we do that, did we think? Well no, obviously it's impossible for Geoff's team to specify tests up front. It just can't be done. No way. But wait... maybe if we did x... or even y? If we'd done that, then we could have avoided the problem that time when the customer forgot zxy... (furrowed brows and discussion follows...)

Brilliant! Someone criticized Geoff's approach and won! We can return to Sweden with lots of new ideas to debate, and a suggestion that could make Geoff's team work even more effectively. Success!

-- EmilyBache

back to Xp2002Stories³OneSuitePractitionersRepor ³³host³³date³June 17, 2002³agent³Mozilla/4.0 (compatible; MSIE 5.0; Windows NT 4.0) Opera 6.03 en?³OneSuitePractitionersReport