CONCEPT Cited by 1 source
Trunk-based development¶
Trunk-based development is the branching model in which a single
long-lived branch (main / master / trunk) is the source of
truth, and every developer merges their small, short-lived feature
branches into it frequently (typically daily or per-PR). Release
branches, if they exist, are cut from main and quickly retired;
they are not a staging area for integration.
The model's load-bearing operational property is that the merge is
the dangerous moment: many people's changes converge on main
continuously, and concurrent merges can produce interaction failures
that each-PR-in-isolation CI can't catch — i.e.
semantic merge conflicts. At
scale, trunk-based development pairs naturally with a
merge queue to defend against this failure
mode.
Contrasted with Git Flow / long-lived-feature-branch models, where integration happens at the release-branch boundary and concurrent-merge pathology is less acute — at the cost of delayed integration + the "integration-hell" merge-heavy end-of-cycle pathology.
(Source: sources/2026-04-29-atlassian-inside-atlassians-merge-queues)
Seen in¶
- sources/2026-04-29-atlassian-inside-atlassians-merge-queues —
Atlassian's Jira repo operates trunk-based (800+ devs, 300+
merges/day into
main); the merge-queue pattern is the natural scale-boundary defence for the model.