1 nota · última: 27 Sep 2013
The post argues that single responsibility works best as a vertical concern tied to feature behavior and state, not as a stack of horizontal layers. It warns that premature separation creates layer soup.