This page collects terms and phrases that appear across essays on this site.
These are not formal definitions.
They reflect how these ideas are used and understood here.
Language shapes how we think. Precision matters.
A system property where different parts converge to the same state over time rather than immediately.
Often treated as a guarantee. In practice, it is a tradeoff that shifts complexity from systems to operations and people.
State that is owned and stored across multiple services rather than a single authoritative source.
Distributed state increases scalability, but weakens clarity about what is currently “true.”
Automatic re-execution of failed operations under the assumption that failures are transient.
Retries hide failure until they amplify it, especially when side effects are not idempotent.
Changes that occur outside the primary intent of an operation, such as sending messages, updating external systems, or triggering downstream actions.
Side effects are easy to duplicate and hard to undo.
A condition where parts of a system succeed while others fail, leaving the overall operation unresolved.
Partial failure is more dangerous than total failure because it appears correct while remaining incomplete.
Clear responsibility for defining, updating, and validating the truth of a system’s data.
When ownership is fragmented, correctness becomes ambiguous even if every component behaves as designed.
Responsibility for declaring whether an operation has succeeded or failed end-to-end.
Without outcome ownership, systems can perform work indefinitely without ever finishing anything.