docs: chart the shortest route to a hand-drivable daemon #415

2026-08-14T23:08:55Z

KofTwentyTwo commented

2026-08-14 23:08:55 +00:00

(Migrated from github.com)

Written after the ADR 0002 sync model landed, to answer one question: what stands between the current engine and a daemon you can drive by hand?

One thing, mostly. The engine executes verified mutations, but Mail has zero mutation RPCs — so nothing can ask it to move, delete or flag a message. Two consequences fall out of the same gap: filters_enabled has no wire surface (settable only by hand-editing SQLite), and placements are invisible to clients.

The doc records:

what works and what does not if you boot dev today, including the rm ~/.nuncio/nuncio_main.db* step that looks like a bug if you hit it cold
the three-step critical path — #407 (in flight) → #395 (the keystone) → a CLI manual-readiness pass
the honest backlog: 81 open issues, i.e. the roadmap through M7, broken down by milestone so the three steps above aren't mistaken for the whole job
known-open items nobody should re-derive, including the recovery::…table_salvage flake that has never reproduced and is not understood
the two test traps that cost real time this month: tracing's process-global callsite Interest cache vs thread-local subscribers, and tokio::time::pause() over real blocking I/O

Docs only.

Written after the ADR 0002 sync model landed, to answer one question: what stands between the current engine and a daemon you can drive by hand? **One thing, mostly.** The engine executes verified mutations, but `Mail` has zero mutation RPCs — so nothing can ask it to move, delete or flag a message. Two consequences fall out of the same gap: `filters_enabled` has no wire surface (settable only by hand-editing SQLite), and placements are invisible to clients. The doc records: - what works and what does not if you boot `dev` today, including the `rm ~/.nuncio/nuncio_main.db*` step that looks like a bug if you hit it cold - the three-step critical path — #407 (in flight) → #395 (the keystone) → a CLI manual-readiness pass - the honest backlog: **81 open issues**, i.e. the roadmap through M7, broken down by milestone so the three steps above aren't mistaken for the whole job - known-open items nobody should re-derive, including the `recovery::…table_salvage` flake that has never reproduced and is **not** understood - the two test traps that cost real time this month: `tracing`'s process-global callsite `Interest` cache vs thread-local subscribers, and `tokio::time::pause()` over real blocking I/O Docs only.

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: KofTwentyTwo/nuncio#415