DRAFT

“We’re not good scientists; we’re not happy when proven wrong.”

- Dr. Aleksandar Dimitrijevic on the psychoanalysis community

Relate “clean code” and “common wisdom” to actual research.

Measure quality. Can’t measure naming, domain.

Hard-to-swallow pill: Logical arguments are not research or proof, only a hypothesis that can potentially (or not) be tested empirically.

If you can’t measure it, you can’t prove it. Come up with new measurements. Accept that some things cannot be measured yet, but maybe in the future.

Too many metrics to consider? Which are common? Because they’re easy to measure, or because they have strong correlated fault density?

  • Fault density
  • Changeability

Fault density is number of bugs per line of code.

Changeability is change request relation to regressions, and how fast a change request is.

Fandom, stardom, authority, ethos, pathos.

The author can claim “I applied this in 100 projects and they all turned out successful”, but without factoring out depending variables, you won’t know if it was in fact this particular technique that caused the success or something else entirely, maybe just having a very good team or stimulating company culture, who knows

Caveats, 50% of bugs can be correlated to company structure? Citation.