Engineering

Boring is a feature: defaults we reach for on every new build.

The unglamorous choices that keep projects maintainable long after launch.

New technology is exciting. Maintaining it three years later usually isn't. On client work and our own products we start from a set of dull, dependable defaults and only move away from them for a clear reason.

Our defaults

  1. One relational database. A well-designed Postgres schema handles far more than most projects need. Add other stores only when there's a measured reason.
  2. Strict types from day one. TypeScript in strict mode catches whole classes of bugs before users do. Turning it on later is painful.
  3. Every schema change in version control. If you can't rebuild your database from the repo, you don't really own it.
  4. Security rules close to the data. Row-level security and server-side checks, never trust the client.
  5. Tests on the money paths. You don't need 100% coverage, but anything involving payments, points or permissions gets tested.
  6. Static where possible. Our own website is plain HTML and CSS. It's fast, cheap to host and hard to break.
  7. Automated deploys. Push to the main branch, it builds and goes live. No manual steps to forget.

Why it matters

We've audited systems that were built quickly with lots of moving parts. The common thread is that they become hard to change safely. Boring defaults make the next change easy, and that's what clients pay for over the life of a product.

Want a health check on a system you've inherited? See our services.

Keep reading

More from TemTed.