From the 2026-08 architecture review follow-up (related to #1781).
What is wrong: "lifecycle" names three different concepts in the domain, so a reader can't tell what a piece solves:
internal/domain/lifecycle — a generic engine registry (Engine[H]: Name/Bind/Start/Health/Stop) used by curators and memory engines; in-memory component lifecycle.
identity.Lifecycle (internal/domain/identity) — an entity status field: active/suspended/retired, driven by suspend/reactivate commands.
storecontract.TaskLifecycle — the store surface of the task state machine (enqueue/claim/transition/requeue), implemented in postgres and guarded by the transition table in internal/taskstate.
Wanted: distinct names per concept (e.g. engine registry stays lifecycle; the task machine is a state machine — taskstate already owns its transition table; identity's is a status), or a recorded decision if the overlap is intentional. No behavior change either way.
Done when:
each concept has one unambiguous name in internal/domain (or a type::decision issue records why they share one)
references updated; task check green
From the 2026-08 architecture review follow-up (related to #1781).
**What is wrong:** "lifecycle" names three different concepts in the domain, so a reader can't tell what a piece solves:
1. `internal/domain/lifecycle` — a generic *engine registry* (`Engine[H]`: Name/Bind/Start/Health/Stop) used by curators and memory engines; in-memory component lifecycle.
2. `identity.Lifecycle` (`internal/domain/identity`) — an entity *status* field: active/suspended/retired, driven by suspend/reactivate commands.
3. `storecontract.TaskLifecycle` — the store surface of the *task state machine* (enqueue/claim/transition/requeue), implemented in postgres and guarded by the transition table in `internal/taskstate`.
**Wanted:** distinct names per concept (e.g. engine registry stays `lifecycle`; the task machine is a state machine — `taskstate` already owns its transition table; identity's is a status), or a recorded decision if the overlap is intentional. No behavior change either way.
**Done when:**
- each concept has one unambiguous name in `internal/domain` (or a `type::decision` issue records why they share one)
- references updated; `task check` green
sam
added the discovery label 2026-11-02 15:23:19 +00:00
Instance found while reviewing: storecontract.TaskPage (Statuses/Workflow/After/Limit) and storecontract.DispatchFilter (BindingID/CaptureID/TaskID/Limit) are the same shape — a list filter plus a limit — but named for different halves of it (page vs filter). In practice TaskPage is a filter: internal/webui/api_tasks.go:97 builds it from URL query params (taskPageFrom(q url.Values)). The proto side calls the dispatch one ListDispatchesRequest and says it "filters". So the asymmetry is accretion, not meaning. Candidate fix under this ticket: one naming scheme, e.g. *Filter + an embedded Page{Limit, After}, or rename TaskPage → TaskFilter. Go-only rename; proto wire names unaffected.
Instance found while reviewing: `storecontract.TaskPage` (Statuses/Workflow/After/Limit) and `storecontract.DispatchFilter` (BindingID/CaptureID/TaskID/Limit) are the same shape — a list filter plus a limit — but named for different halves of it (page vs filter). In practice TaskPage *is* a filter: internal/webui/api_tasks.go:97 builds it from URL query params (`taskPageFrom(q url.Values)`). The proto side calls the dispatch one `ListDispatchesRequest` and says it "filters". So the asymmetry is accretion, not meaning. Candidate fix under this ticket: one naming scheme, e.g. `*Filter` + an embedded `Page{Limit, After}`, or rename TaskPage → TaskFilter. Go-only rename; proto wire names unaffected.
sam
added the medium label 2026-11-14 14:45:43 +00:00
sam
added the deferred label 2026-11-15 17:26:12 +00:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
From the 2026-08 architecture review follow-up (related to #1781).
What is wrong: "lifecycle" names three different concepts in the domain, so a reader can't tell what a piece solves:
internal/domain/lifecycle— a generic engine registry (Engine[H]: Name/Bind/Start/Health/Stop) used by curators and memory engines; in-memory component lifecycle.identity.Lifecycle(internal/domain/identity) — an entity status field: active/suspended/retired, driven by suspend/reactivate commands.storecontract.TaskLifecycle— the store surface of the task state machine (enqueue/claim/transition/requeue), implemented in postgres and guarded by the transition table ininternal/taskstate.Wanted: distinct names per concept (e.g. engine registry stays
lifecycle; the task machine is a state machine —taskstatealready owns its transition table; identity's is a status), or a recorded decision if the overlap is intentional. No behavior change either way.Done when:
internal/domain(or atype::decisionissue records why they share one)task checkgreenInstance found while reviewing:
storecontract.TaskPage(Statuses/Workflow/After/Limit) andstorecontract.DispatchFilter(BindingID/CaptureID/TaskID/Limit) are the same shape — a list filter plus a limit — but named for different halves of it (page vs filter). In practice TaskPage is a filter: internal/webui/api_tasks.go:97 builds it from URL query params (taskPageFrom(q url.Values)). The proto side calls the dispatch oneListDispatchesRequestand says it "filters". So the asymmetry is accretion, not meaning. Candidate fix under this ticket: one naming scheme, e.g.*Filter+ an embeddedPage{Limit, After}, or rename TaskPage → TaskFilter. Go-only rename; proto wire names unaffected.