Some of the suggestions and points in the article that are of note include:
- 90% of scientists are self-taught programmers
- All aspects of software development should be broken into tasks roughly an hour long
- Provenance of data refers to data that is accompanied by a detailed list of code and operations for recreating the data and code output
- Programmers should work in small steps with frequent feedback and course corrections
- Use assertions (executable documentation) to avoid mistakes in code
- Scientists should reprogram complicated tasks to make them simpler for a human to read instead of including paragraphs of comments explaining how the code works.
While largely approachable, the paper still suffers from a slight overuse of jargon from the software development field. As a result, the importance of some of their recommendations escapes me.