Architecture · 10 min
Architecture decisions that survive a decade
Most architecture is wrong about the future. A discipline for making decisions that age well, and for knowing which decisions are worth making permanent.
The failure mode of architecture is over-committing to predictions about a future that does not arrive. We design for the system we imagine, and the business outgrows it in a direction we did not foresee.
The discipline is to separate reversible decisions from irreversible ones, and to spend the budget of certainty only where it matters. Most decisions should be made cheaply, documented, and reversible. A few — data ownership, boundaries, identity — deserve permanence.
Boundaries are the decisions that matter most. Get the domain boundaries right and the rest is refactorable. Get them wrong and every later decision is a compromise against the original mistake.
Document decisions as they are made. The Architecture Decision Record is the most undervalued artifact in engineering. It is how a system stays coherent across the people who will inherit it.