Implementing the Holbrook Limit: Modules

The best way to respect the Holbrook Limit is to explicitly divide the system codebase into units, typically stored as a diskfile, and to mechanically verify at compile time that they obey the limit.

Modern languages such as SML refer to such units as modules, and provide an explicit support infrastructure for them called the module system. In SML, the module system comprises about one-half the complexity of the language and implementation. The module system makes syntactically and semantically explicit:

• Just which lines of code comprise the module itself.
• The exported interface -- what parts of the module are externally visible.
• The imported interfaces -- what other modules the code needs to function.

The psychological complexity of a module is proportional to the number of lines of code in the module itself, plus the psychological complexity of the imported interfaces.

To keep a system piecewise comprehensible as a whole, the psychological complexity of each individual module must be below the Holbrook Limit.

Back to Comprehensible Computing.