My post on DZone about Humble Architects sparked somewhat of a controversy, especially around my disparaging comments regarding Spring and Dependency Injection Frameworks. In this post, I expand on why I stopped using Spring.
I was one of the earliest adopter of Spring in Norway. We developed a large system where we eventually had to start thinking about things like different mechanisms for reuse of XML configuration. Eventually, this evolved into the @Autowire and component-scan which took away the problem with huge configuration files, but in return reduced the ability to reason about the whole source code – instead isolating developers in a very small island in the application.
The applications tended to blossom in complexity as either the culture, the tool, the documentation or something else made developers build unnecessary layer upon unnecessary layer.
Later, I tried to build applications without a Dependency Injection framework, but taking with me the lessons about when to use “new”, when to have a setter or a constructor argument and which types were good to use as dependencies and which created coupling to infrastructure.
So I found that some of the instincts that the DI container had given me made me improve the design, but at the same time, I found that when I removed the container, the solution became smaller (which is good!), easier to navigate and understand and easier to test.
This leaves me with a dilemma. I found that the cost of using the container is very high – it creates a force towards increasing complexity and size and reduced coherence. But at the same time, it taught me some good design skills as well.
In the end, creating a coherent, small system is to me of much higher value than to create one that is decoupled just for the sake of decoupling. Coherence and decoupling are opposing forces and I side with coherence.
At the same time, I found that the culture around dependency injection has a very strong preference for reuse. But reuse does introduce coupling. If module A reuses module B, module A and B are coupled. A change in B might affect A for better (a bug fix) or worse (an introduced bug). If the savings from the reuse are high, this is a trade-off worth making. If the savings are low – it is not.
So reuse and decoupling are opposing forces. I find myself siding with decoupling.
When there is a conflict, I value coherence over decoupling and decoupling over reuse. The culture where Spring is in use seems to have opposite values.