a little madness

Earlier today I read Cedric Beust’s post Agile People Still Don’t Get It. In that post Cedric makes some fair criticisms of misguided agile evangelism. I find it particulalry ironic that the closing sentence is:

So here is my advice to Agilists: get real now, or you will become irrelevant soon.

Pragmatism should be a cornerstone argument for agile development, so those evangelists that don’t emphasise it surely are doing more harm than good¹.

Having said all that, there is an interesting undertone in some of Cedric’s comments that got me thinking. In his defense of pragmatism, Cedric makes repeated assertions that sometimes we need to cut corners to make a real-world deadline. One corner to be cut is testing:

But the truth is: at times, I don’t do TDD because implementing a feature quickly is more important than a fuzzy feeling.

This is what really got me thinking. Note that we are talking short-term here: in the long term, I think most would agree that cutting corners will catch up with us eventually. But why do we assume that cutting this corner will save time in the short term? I think the assumption is based purely on a shallow evaluation: writing less code this second means I can hit the looming deadline more easliy. This seems valid enough: writing the extra code to perform the testing takes non-zero time after all. But the analysis is missing something: can we really implement a feature without testing it? We can try, but how will we know if it works at all? A broken feature is worse than no feature at all, so there is no way to invoke the “customer value” argument here.

I think most people agree that some level of testing is required: the risk of zero testing is just too great. A temptation, then, is to test the feature by hand, because that will be quicker than writing code to test it. Now the assumption is getting even shakier. Maybe a manual test is quicker as a once-off, but what about when you find a problem? Now you need to fix the code, then manually test again. You will be wishing you had just written test code in the first place, so you could repeat the test with very little extra effort. So, counter-intuitively, it may be quicker to code the feature and tests rather than just the feature. Not in all cases, for sure², but probably more frequently than we realise.

In fairness to Cedric, I have taken this off on a tangent of my own. His post just got me thinking about a natural assumption that we tend to make which may actually be false. I agree with the assertion that sometimes a pragmatist has to make compromises to meet a short term goal, I just wonder about the short term effects of this particular compromise.

—
¹ I suppose this is what happens when people become true believers in a methodology: they start to drift farther and farther from reality.

² For example, when the feature is very hard to test in an automated fashion. Usually this can be remedied in the long term, when you can afford the extra time to refactor for testability.

This entry was posted on Thursday, June 8th, 2006 at 5:51 pm and is filed under Agile, Technology, Testing. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.