Bertrand Serlet stopped by Stripe yesterday to do a fireside-chat style interview with Patrick. Bertrand’s a well-traveled former SVP of Apple, veteran of PARC and NeXT, and quite a lively and entertaining speaker.
Among the ideas that he expressed, one of my favorites was that software needs a certain amount of constant “anti-entropy maintenance”. Left alone, it’ll tend to get slower over time as new features and inefficient bug fixes are introduced, and this builds toward death by a thousand cuts. For every big OSX release, Apple committed a baseline level of engineers to fight the effect, along with the normal pool of engineers to do feature work.
This couldn’t have rung more true with me; all software starts out good, and all software ends up bad 1. The best we can do is maintain our projects to the highest possible standard in the interim to stave off that fall into darkness as long as possible. Some organizations never acknowledge the problem, some implement periodic clean-up phases that attempt to improve code quality retroactively, but the best possible model is the one that Bertrand suggest, where a permanent staffing of engineers to apply constant backpressures to keep the overwhelming entropic forces at bay.
1 That’s “all” with a few exceptions; most notably, OSS projects that can afford to commit a disproportionate amount of time in the interest of archieving the highest quality standards. But this level of expensive attention to detail is vanishingly rare within a resource-constrained corporate environment.
Did I make a mistake? Please consider sending a pull request.