Staying sane while writing software

This is a collection of principles, techniques, and paradigms that lead to useful, quality software while preserving the sanity of the programmers writing it. Mental health and project success go hand-in-hand.

Not all of the principles mentioned here can hold in all cases. Some may conflict in a particular setting. But where possible, applying these ideas seems to lead to the most satisfying results.

Currently this is just an outline.

Attributes of desirable software

What most people (should) want

• features!
• customization
• privacy
• security
• robustness
• performance
  • responsiveness
  • low resource consumption

What programmers also want

• repeatable behavior
• incremental effects
• local reasoning
• easy rearrangement
• generalization
• efficiency
• scalability
• consistency

Qualities that help satisfy these desires

Relatively abstract

• monotonicity
  • accumulation described by sets or, more generally, lattices
• idempotence
• commutativity
  • The Scalable Commutativity Rule
    • "Whenever interface operations commute, they can be implemented in a way that scales."
• associativity
• orthogonality
• composability

Somewhat more concrete

• principle of least authority
• declarative description
• redundancy
• tolerance
• isolation

Some techniques that can provide these properties

Broad concepts

• modular design
• decentralized authority
  • distribution of resources, labor and responsibility
• granular delegation of authority
  • as in the object-capability model
• clear, incorruptible sources of truth

More specific architectural choices

• purely functional cores with imperative shells
  • Boundaries
• shared global state without local state
  • Local State is Poison
  • Local state is harmful
• object capabilities
  • as in the E Programming Language
• caching
• replication
• append-only databases
  • event sourcing
  • the log as a unifying data abstraction
  • Beating the CAP Theorem
• consistent or rendezvous hashing
• SOA (in the microservices sense)
  • smart endpoints, dumb pipes
  • service failure resilience
• REST (in the hypermedia-driven sense)
• CRDTs
  • Conflict-free Replicated Data Types
  • Strong eventual consistency
• LVars
  • Lattice-based Data Structures
• CALM
  • Consistency as logical monotonicity
  • as in the Bloom programming language
• Relational Programming
  • Programs that can be run in multiple directions
  • as in the miniKanren EDSL
• Probabilistic Programming
  • Programs that describe and perform inference in probabilistic models
  • as used in Probabilistic Models of Cognition
  • The MIT Probabilistic Computing Project