Notes for chapter 5: "Taming complexity"

Errors in scientific software

A few recent articles have pointed out or illustrated the problem of errors in scientific software:

• The Nature News report Computational science: ...Error by Zeeya Merali was one of the first articles to draw attention to the problem.
• Rampant software errors may undermine scientific results by David Soergel explains on theoretical grounds why the problem may be more important than most scientists suspect.
• Are all global alignment algorithms and implementations correct? by Tomáš Flouri et al. illustrates in a careful case study that even old widely used computational methods and their implementations can be wrong.
• The Effects of FreeSurfer Version, Workstation Type, and Macintosh Operating System Version on Anatomical Volume and Cortical Thickness Measurements by Ed Gronenschild et al. emphasises the complexity of computational environments and software as a cause of variations in scientific results.

Formal verification of software

The first part of a lecture by Benjamin Pierce (47 mn) gives a good overview of how specifications are used today in real-life software development. The second part is about a specification for the Dropbox synchronization service.

Managing state

A short and pragmatic discussion of why functional programming matters for dealing with program complexity is given in a blog post by Krys Jenkins.