Revisiting the Dependable Dependencies Principle

Twelvety-three years ago I published my Dependable Dependencies Principle for modular software design. Over the years, I’ve expanded on it, but never written that intervening thinking down. So here we are. It’s a pretty straightforward thesis: the more depended-upon a module is, the more dependable it needs to be. This gives one dimension we can use to assess the potential risk that modules present in terms of the “blast radius” when they fail – how much stuff could that impact? Empirical research does indeed find a correlation between how depended-upon a module is and the severity of bugs traced to that module, so we’re on solid ground here. Metrics like Page Rank (named after Larry Page of Google, who invented the algorithm used to rank the importance of web pages based on the links between them) can be a usable proxy to centrality when we apply them to modules and their dependencies. The highest-ranked modules in the system are either directly or indirectly depended on the most –…