Favor Outcomes Over Rules
Cargo Cults and Angry Monkeys
If you have never heard the term “Cargo Cult Programming“, it’s an it’s an excellent way to describe blindly following processes without understanding the theory behind them. The term originates from a story about aboriginal islanders during World War II. During the war, cargo planes regularly (and from their perspective, magically) arrived with supplies and food that benefited these islanders and then after the war the planes stopped coming. This apparently didn’t stop the aboriginals from building ad-hoc landing strips and the like in hopes of ‘summoning’ more planes with supplies. If that story isn’t to your liking, there is another one of questionable accuracy about monkeys and bananas.
I’ve made a fair number of posts lately that address subjects near and dear to programmers while still having broader reach and this is certainly another such subject. Doing things without understand why is generally the province of the incurious or the busy, but I think it’s worth generally forcing yourself to ask why you’re doing pretty much anything, particularly if you’re in a fairly cerebral line of work. As professionals, it behooves us to realize that we want supplies or that we want not to get sprayed rather than thinking that building landing strips and beating up monkeys are just things that we do and such is life.
In fact, I’d venture to say that we pride ourselves in this sort of thing as we advance in our careers. As programmers, we even help our users understand this concept. “I understand that you normally push the green button three times and hit the backspace key while pressing the mute button, but what are you actually trying to do when you do all that?” How often have you asked a user something like this? In essence you’re saying “forget the process for a minute — what’s the larger goal here?”
Do Unto Others
It then bears asking, if we pride ourselves on critical thinking and we seek to help users toward that end, why do we seem to encourage each other to dance for supplies and spray our team members with cold water? Why do we issue cryptic cargo cult orders to other programmers when we understand our own rationale? Let me give an example.
A few years back I was setting up a new machine for work on a project to which I was new, and the person guiding me through the setup told me that I needed to install KDiff, a spectacularly average comparison tool that even then hadn’t been revved or maintained in a few years. Now, I have nothing, per se, against KDiff, and the price is right, but this struck me as a very… specific… order. What difference did it make what I used for diff? I had used a paid version of Beyond Compare prior to that and always liked it, but when it comes to tooling I do enjoy poking around and so I didn’t mind trying a different tool.
I’m pickier about some things than others, so I just said “Bob’s Your Uncle,” installed KDiff and didn’t think about it again for a long time. It was actually many months later that I was sitting in on a code review for a developer who had just started on the project and found out the reasoning behind this cargo cult practice. The developer whose code was being reviewed casually opened BeyondCompare and one of the more tenured reviewers than me promptly freaked out that he wasn’t using KDiff. Bemused, I tuned in to hear the answer to a question that I’d completely forgotten about. The issue is that the project had a handful of XML files containing application meta-data that were source controlled and that some unnamed developer had, at some point, done something, presumably with some sort of diff tool, that had allegedly hosed one of these files by screwing up the text encoding. Nobody in the room, including the freaker-outer, knew who this was or exactly how or when it had happened, but nonetheless, it had been written in stone that from that day, Thou Shalt Use KDiff if you want to avoid punishment in the form of Biblical plagues.
This certainly wasn’t my only experience with such a thing. More recently, I overheard that the procedure for editing a particular file in source control was to grab the latest, edit it really fast and then check it in immediately. I was a bit perplexed by this until I learned that the goal was to avoid merge conflicts as this was a commonly edited file. I thought, why not just say in the first place that a lot of people edit this file and that merge conflicts are bad, in the same way that I had previously thought “why not just tell people that you’re worried about messed up encodings?”
And that line of questioning really drives to the heart of the matter here. Developers are generally quite skilled and pretty intelligent people and, more importantly, they tend to be be people that like to solve problems. So why not give them the benefit of the doubt when you’re working with them? Instead of giving your peers rote procedures to follow because they happened to work for you once, why not explain the problem and say “this is how I solved it, but if you have a better idea, I’m all ears!”
And you know what? They might. What if instead of forcing people to use KDiff, someone had written a validation script for the offending file that prevented checkins of a badly formed edit? What if instead of having flash edit-checkins of a file, the design of the application were altered to eliminate the contention and potential for error around that file? I suggest these things because I’m a fan of making the bad impossible. Are they the greatest ideas? Are they even practical in those situations? I honestly don’t know, but they might be approaches that have the advantage of placing fewer restrictions on developers or meaning one less thing to remember.
I would encourage you to trust your peers to grasp the bigger picture whenever possible (unless perhaps they’ve demonstrated that this trust isn’t warranted). As I say in the title, it seems like a good idea to favor describing desired outcomes over inventing rules to be memorized. With the former approach you free people you work with to be creative. With the latter approach, you constrain them and possibly even stress them out or frustrate them. And who knows, they may even surprise you with ideas and solutions that make your life easier.
Great post. It’s helpful to remember that computers run programs, people are supposed to think. If your job can be done by following a set of instructions without the need for thought, then you’re probably a candidate to be replaced by automation.
Thanks — glad you liked the post. And the bit about candidates being replaced by automation combined with the theme of this post reminds me of this demotivator for some reason: http://despair.com/motivation.html