Agile development has created a culture of newly weds, programmers coupled in pairs oblivious to the fate that awaits them. As with all forms of coupling, the short-term benefits are outweighed by the long-term consequences. The optimism of a new relationship spelled out in code never lives up to the story, no matter how it is prioritised.
There has been much talk and many studies about how effective pair programming is, but clearly all those involved are looking for some kind of meaningful justification that makes sense of their predicament. Apparently pairing improves code quality and is enjoyable, but I doubt that: how can you really have fun and program well when you keep having to remove your headphones to listen to someone else questioning your mastery of code?
Good pairing is supposed to involve alternately navigating and driving. From what I can tell, this means navigating the quirks of another’s style and conventions while driving home your own beliefs about how to organise things properly. It is a contest in which there will be a winner and a loser. So much for team spirit!
I suspect that financial debt – which is like technical debt but with money – is a contributory factor. PCs, however, are not that expensive. Surely companies can spare enough money to supply each programmer with their own PC or, at the very least, a keyboard they can call their own? The point of a PC is that it’s personal – the clue is in the name. Sharing a computer is like sharing a toothbrush, only more salacious.
For example, the practice of promiscuous pairing is often promoted.
You swing from one partner to the next willingly, openly and frequently. Such loose coupling demonstrates a lack of commitment and sends out the wrong moral message. If you’re going to have to pair, you should do it properly, all the way from ‘I do’ to ‘Done’. It is likely that there will be an eventual be a separation of concerns, but that at least avoids the risk of communicating state-transition diagrams and infecting your C++ code with explicit use of the standard library namespace.
One thing that might be said in favour of pairing is picking up new skills. For example, I have learnt to use a Dvorak keyboard and a succession of editors with obscure key bindings and shortcuts. Being able to present new and existing partners with an unfamiliar and hostile environment puts them off their guard and sends out a clear signal about the roles in the relationship. I also find pairing can be effective with newbies. They can either sit and watch for a few hours or they can drive while you correct them from the back seat.
These benefits, however, are few and far between. The day-to-day reality is more cynical: the constant nagging, the compromises you make, the excuses you have to make up, the methods you use, the arguments, the rows, the columns... and you sometimes have to put up with your partner snoring after you’ve offered an extended and enlightening explanation of some minor coding nuance they seemed apparently unaware of!
So, don’t impair to code, decouple.
Overload Journal #102 - April 2011 + Process Topics
|Browse in :||
All > Topics > Process (52)
Any of these categories - All of these categories