Not Quite Modern
Reviewed by Greg Wilson / 2022-05-06
Keywords: Book Review
Despite its title, David Farley's new book Modern Software Engineering feels a bit old-fashioned. The author is best known for his advocacy of dev ops and continuous delivery, and his starting point is the following definition (pg 17):
Engineering is the application of an empirical, scientific approach to finding efficient, economic solutions to practical problems.
I strongly agree that our field needs more of this; I was therefore disappointed that Farley proceeds to leave out almost everything that empirical, scientific study of programs, programming, and programmers has uncovered in the last three decades. As just one example, consider test-driven development. On pg 98, Farley writes:
There are several studies, academic and informal, on the impact of TDD on defect reduction. Most studies agree that defect reduction is in a range from 40 percent to well over 250 percent.
He includes these three links to articles from 2003, 2007, and 2012 respectively. What he doesn't include is Turhan et al's meta-analysis in 2011's Making Software, which found that (a) the more rigorous the study, the weaker the evidence for or against TDD and (b) overall, there is no evidence that TDD makes anything better or worse. There's also no mention of the more recent (and very careful) work of Fucci et al. (which we reviewed in 2016), which found that (a) there's no difference between test-first and test-after but (b) interleaving short bursts of coding and testing is more effective than working in longer iterations. If people started working in shorter cycles when they adopted TDD they might well ascribe any benefits they experience to the wrong cause; that's exactly the sort of question careful study could answer.
I went through the footnotes in this book twice looking for any mentions of research findings from the last ten years. I came up empty, and I think that's a shame. I'm the first to admit1 that most academic studies in software engineering are irrelevant to most working programmers, but there are a lot of gems. We know so much about how programmers actually do code reviews, whether code clones are actually harmful, what difference static typing actually makes, and where hiring and onboarding go wrong; I think a modern introduction to software engineering should include all of that and more.
- "Admit" is putting it mildly: "shout repeatedly" is probably a more accurate description of how often and how loudly I've complained about this.