A new edition of ThoughtWorks’ Technology Radar was published this week.
To no one’s surprise, AI is the big theme of this edition, I was however attracted by 2 items:
Mermaid is in the adoption quadrant.
In my current project at ING, we used documentation-as-code from the beginning, except that our choice is PlantUML. What I can comment on my experience, in 3 aspects:
- Consistency: By treating documentation as code, it becomes easier to maintain consistency between the code and its documentation, especially when it comes to software architecture. We use C4 and when we make structural changes, the diagrams and the new code are versioned in the same PR, showing exactly the evolution of the system;
- Collaboration: Between engineers is great because it sounds more natural to edit documentation in the same flow as the software is edited. Outside the world of engineers, there is a barrier as it is necessary to know the markup and a point-and-click interface is more intuitive. This is evident in the discussion of the software context, in which interaction with business colleagues is necessary;
- Automation: This is where documentation-as-code shines. If the team culture encourages comments on your codebase, several diagrams can be automatically generated.
It’s difficult for me to agree with the argument that we should embrace complexity in software development. Complexity is something that must be combated in our design, in our implementation, in our processes. Using Cynefin as guide, our goal is to transition from the complex to the complicated and, look, the article uses AI as an example of complexity but AI operates in the complicated, by using patterns and knowledge to deliver decisions and answers.
I remember Dumbledore saying to Harry: “Soon we must all face the choice, between what is right and what is easy”. An architect’s job, and all engineers are architects to some extent, is to resist the temptation of the easiest solution; It is common in these situations to introduce accidental complexity. If we are talking about essential complexity, ok, but we should still fight to reduce it. But I don’t blame Thoughtworks for the approach, Dijkstra once said “complexity sells better”.
Read the Radar, it’s always interesting, especially check out what’s close to the adoption level.