Trunk-based development and short-lived branches

What it is

Trunk-based development means that all developers integrate into a single shared branch ("trunk", usually main) frequently, typically at least once a day. Branches, where used, live for hours or a few days and are merged as soon as the change passes the build. Larger changes are made in small steps that keep the trunk releasable, using feature flags or branch-by-abstraction to hide incomplete work.

Why it matters

Fowler's description of continuous integration rests on frequent integration to a mainline with a self-testing build; long-lived branches defer the integration and hide conflicts until they are expensive. Short-lived branches keep review scope small and let the pipeline verify the actual combined state rather than isolated work.

How to apply

• Break work into steps that can each be merged without breaking the trunk.
• Guard unfinished behaviour with a flag rather than a branch.
• Run the full build and tests on every merge; treat a red trunk as the team's first priority.
• Cut releases from the trunk by tag; use release branches only for supported older versions.

Pitfalls

Without a fast, reliable build the model collapses into "everybody breaks main". Flags that are never removed become permanent complexity. Teams new to the approach often merge too rarely rather than too often.