CEWT is the Cambridge Exploratory Workshop on Testing, and for its fifth meeting, hosted at Linguamatics, a bunch of local testers gathered to consider a question: Theory Over Practice or Practice Over Theory?
A couple of months before CEWT #5 it was posed.
For a day during CEWT #5 it was explored.
Immediately after CEWT #5 it was .... still mostly undecided.
But along the way we did at least have a unicycle, some Transformers, an Alien Dance Party, a selection of retro Nokia phones, and Batman. Oh, and Lean Coffee, and a retrospective, and six talks.
First up in the talks Karo Stoltzenburg tested the question. Taking Weinberg and Gause's classic Are Your Lights On? as her guide, she wondered whether there was a problem here and, if so, whose problem it was, how it got to be a problem, and whether or not it was worth solving. Noting the potential for ambiguity in key terms in the question and the Call For Participation, she carefully teased out possible interpretations and asked us to consider ways in which they might be interesting, scenarios in which they might be applicable, and whether we had to choose between theory and practice at all.
Next, Sneha Bhat and me offered the idea that there's really no opposition between theory and practice, except when labels such as theorist or practitioner are in play. We defined theory as "data that we care to keep track of", and suggested that a sensible way to proceed in a task is to start with the data that you think you need to accomplish it. If you have what you need, then theory has provided the answer. If you don't then you need to practice to get more data. Practice can be guided by theory — because practitioners change their actions based on what they know — and so we have a loop: practice generates data, data becomes theory, theory guides practice.
A transition from waterfall to agile processes at a mobile phone company across around 10 years and six device iterations was the topic of Milosz Wasilewsk's experience report. The theory here was that agile would be faster and better, although in ways that management failed to transmit sufficiently clearly to the staff. The reality was that in practice, a working system — at least one that was reliably delivering product on time — got broken, with multiple late and aborted releases. Milosz's observation was that the theory was good, but the practice didn't live up to it. For example, testing approaches didn't really change and there was little appetite for it to change. Understanding the reasons for the difference between theory and practice is complicated by the number of variables which here include the change in volume of production, the change in the size and ethos of the company, and the change in the competitive landscape over the time period.
In his talk, Aleksander Simic tried to identify the triggers that cause him to switch from theory (reflection, research) to practice (interacting with the system under test). He sketched out on the whiteboard how joining a already-running project team had been challenging in multiple ways including new people, new code base, new architecture, new tooling, and a dramatic difference in scale between the test and production environments. With a lot to learn he found himself constantly discovering that what he was seeing wasn't caused by what he thought, for example after running an experiment and finding some anomalous results which he attributed to the product, it turned out that the hardware configuration had been changed. His testing theory was compromised by constraints that he was unaware of. Discovering those constraints forced him to stop testing (practice) and start understanding the constraints (theory).
Alan Wallace presented an analysis of an analogy between professional swimming, testing, and other professions. He introduced the term training to the discussion and enumerated factors of it, such as that it promotes System 1 thinking, that it can prepare the trainee for performance under pressure or in new scenarios, that dedication to it comes from understanding the value of it, and that it builds competence in the relevant activity. Having done this, he tried to compare professions and found that there seem to be those that are biased significantly in favour of training over execution (e.g. astronaut) and those that are not, or that perhaps tend to train during execution (e.g. testing). He noted that there seem to be few in between.
Finishing us off, Neil Younger reflected on how learning to ride his mountain unicycle had been a mixture of theory and practice, and that reflecting on it has given him ideas about how he might guide others when learning to ride, or to test. He noted how the choice of theory or practice can be dictated by the context: at work he will mostly practice testing, and at home he'll read up on theory, in part through necessities such as availability of hardware, level of interruption, and business needs. He described how initially failing at practice (he took months to learn to mount his unicycle) can be avoided with short cuts from theory (he found a YouTube video suggesting that learners put a house brick behind the wheel) but that he felt this would have deprived him of other learning, such as balancing while stationery. He wondered whether there are short cuts to testing learning and what the trade-offs might be.
This question was picked up in Lean Coffee, where we tried to unravel some of the threads from the day, along with two others:
- Can we identify testing short cuts?
- Is thinking practice?
- Does it make sense to talk about best theories?