Foundation / checks (client) (push) Successful in 4s
Foundation / checks (checks) (push) Successful in 4s
Foundation / checks (deploy) (push) Successful in 4s
Foundation / checks (ecs) (push) Successful in 4s
Foundation / checks (network) (push) Successful in 4s
Foundation / checks (postgres) (push) Successful in 4s
Foundation / checks (tsan) (push) Successful in 4s
Foundation / postgres-tools (push) Successful in 5s
Foundation / image (push) Successful in 5s
Foundation / server-image (push) Successful in 5s
112 lines
7.8 KiB
Markdown
112 lines
7.8 KiB
Markdown
# M11.3: Primitive locomotion, feeding and predation
|
|
|
|
This gate turns the funded, persistent primitive founder from a stationary life
|
|
record into an organism that can act in a finite cellular habitat. The target is
|
|
the exploratory hands-on check defined in [ROADMAP.md](../ROADMAP.md): after
|
|
11.3 and before expanding mutation in 11.4. It is not a balance sign-off.
|
|
|
|
## Authority and boundaries
|
|
|
|
Reviewed PROJECT_VISION, ENGINEERING_CONSTITUTION, SYSTEM_OVERVIEW,
|
|
SIMULATION_ARCHITECTURE, REGION_DURABILITY, NETWORK_ARCHITECTURE,
|
|
PERSISTENCE_ARCHITECTURE, ENERGY_MODEL, ENVIRONMENTAL_PROCESS,
|
|
ENVIRONMENTAL_TRANSFERS, REPRODUCTION_MODEL, and ADRs 0031, 0032, 0053,
|
|
0089, 0098 and 0100. Inspected the public `PrimitiveLives`,
|
|
`PrimitiveRegionHost`, `RegionHost`, `EnvironmentalTransfers`, metabolism,
|
|
development, checkpoint and action-record interfaces and their lifecycle,
|
|
transfer, QUIC and process tests.
|
|
|
|
The M6 `EnvironmentalCommunity` and `EnvironmentalTransfers` remain bounded
|
|
fixtures. They do not become gameplay state, biological policy, or an implicit
|
|
source of nutrients. The simulation and biology libraries remain independent
|
|
of transport, PostgreSQL, graphics and Kubernetes. `evo-server` owns I/O,
|
|
durable admission and scheduling; deterministic biological resolution receives
|
|
explicit state and intentions.
|
|
|
|
ADR-0098 already selects movement and engulf intentions, phenotype-derived
|
|
maintenance/movement/intake, channel-by-channel absorption, and conserved
|
|
engulfment/corpses. This plan does not choose additional gameplay constants.
|
|
Starting thresholds marked tunable stay versioned and visible to playtesters.
|
|
|
|
## Contract gaps to close
|
|
|
|
1. **Persistent progression:** checkpoint profile 3 currently closes only
|
|
creation state. 11.3 must define a new versioned cellular checkpoint closure
|
|
for position, tick, organism/genome, body/reserve, finite habitat fields,
|
|
action/retry state and all implemented biological timers. Recovery must never
|
|
supply defaults or advance biology while offline. Preserve profile-3 bytes and
|
|
genesis compatibility.
|
|
2. **Atomic actions:** the existing durable action record admits only the M10
|
|
no-effect probe kind, and its batch trace belongs to the environmental
|
|
fixture. Define a versioned cellular intent and publication path that records
|
|
the checkpoint and final outcomes together. Reuse the existing lease,
|
|
control-generation and exact-retry guarantees; never duplicate an environmental
|
|
debit or organism effect after an uncertain commit.
|
|
3. **Spatial habitat and perception:** the shared world voxel pitch is 1 m
|
|
(`cell_pitch_mm = 1,000 mm`), while the founder's body extent is 20 μm.
|
|
M6 `HabitatBox` supports micrometre extents inside a voxel, but `FieldStore`
|
|
permits only one site per voxel and environmental views sample that whole
|
|
site's stocks. It cannot represent two independently stocked sub-meter food
|
|
patches in one voxel. A cell could cross 50,000 body widths before reaching
|
|
the next voxel, so wiring feeding directly to those field sites would make
|
|
steering and local food choice meaningless. This is an architectural scale
|
|
gap; do not change the shared world voxel contract to hide it.
|
|
|
|
**Resolution:** [ADR-0101](../adr/0101-cellular-microhabitat-resources.md)
|
|
selects a new sparse, bounded two-dimensional microhabitat, with independently
|
|
stocked 1,000 μm sites (50 founder extents), explicit finite per-channel
|
|
openings and retained provenance. The pitch is a versioned tunable, not a
|
|
final world rule. Its opening is provisioned as part of the region; it has no
|
|
automatic replenishment. Later coupling to meter-scale world stores must be an
|
|
explicit conserved transfer between scales. [ADR-0102](../adr/0102-cellular-spatial-trial.md)
|
|
now defines bounded non-wrapping motion and footprint contact. The cellular
|
|
profile must still define perception range and resource visibility. Do not
|
|
use the probe marker as an organism or expose unseen resources as client
|
|
knowledge.
|
|
4. **Owner routing:** 11.2 still lacks authenticated routing/reconnect when an
|
|
account's life is in another region. Resolve the ADR-0100 protocol gap (v6
|
|
owning-region response plus authoritative endpoint lookup, or server-side
|
|
routing) before claiming multi-region life control. A single-region local
|
|
playtest may be prepared independently, but cannot close 11.2 or validate
|
|
cross-region play.
|
|
|
|
No new dependency or renderer is required. If the selected spatial, checkpoint,
|
|
or publication design conflicts with accepted architecture, stop that affected
|
|
implementation and resolve the contract first.
|
|
|
|
## Delivery segments
|
|
|
|
For each segment, name affected modules, run its applicable tests and checks,
|
|
record unavailable environments in `VALIDATION.md`, then commit and push the
|
|
scoped change before starting the next segment.
|
|
|
|
| Segment | Affected modules | Acceptance evidence |
|
|
|---|---|---|
|
|
| 11.3a: cellular state and spatial contract | Biology public values; server lifecycle; checkpoint profile; schemas and architecture docs | Implement ADR-0101/0102 microhabitat, movement and contact; define perception range, resource visibility, units, tick order and version allowlist; bounded canonical codec; hostile/truncated inputs refused before allocation; profile-3 compatibility; exact restore of every state field |
|
|
| 11.3b: server-authoritative movement and perception | Biology adapter; spatial query; intent validation; network protocol/client session | Clients send only bounded movement/engulf intentions; server decides speed, position, contact and visible environment; range/speed forgery and stale intent tests; deterministic replay |
|
|
| 11.3c: metabolism and finite transfers | Biology metabolism/development adapters; cellular field ledger; corpse/predation resolution | Intake and costs derived from phenotype; each resource channel conserves independently; contested stock has one winner; engulfed inventory moves once with excess detritus; starvation and death preserve all inventory |
|
|
| 11.3d: durable tick publication and recovery | Primitive region host; persistence adapter/schema; session action outcomes; process tests | Checkpoint, habitat changes, terminal outcome and final action results publish atomically; exact retry after uncertain commit; kill/restart at publication boundaries; no offline progress; Linux/PostgreSQL CI qualification |
|
|
| 11.3e: exploratory playtest slice | Server view replication; existing engineering client input/view; playtest notes | One player can spawn, see what the cell perceives, steer, seek and feed in a finite habitat; show movement/resource costs and failure states; capture reproducible trace and human notes; no mutation or reproduction implied |
|
|
|
|
## Required checks
|
|
|
|
- Unit/property coverage for position bounds, spatial contact, intention
|
|
ordering, metabolism, transfer conservation and adversarial inputs.
|
|
- Deterministic multi-seed traces, simultaneous resource contention, speed/range
|
|
forgery, starvation, engulfment, corpse recovery and closed-box conservation.
|
|
- Canonical encode/restore under `NDEBUG`, old profile/golden regressions,
|
|
Linux ASan+UBSan, clang-format, clang-tidy and applicable Windows network tests.
|
|
- Real PostgreSQL process tests for uncertain publication, SIGKILL and restart;
|
|
CI evidence is required where the local Windows environment cannot run them.
|
|
- A short hands-on probe before 11.4 that records agency, confusion, useful
|
|
feeding choices, cost legibility and reasons to continue. One session is
|
|
exploratory feedback, not evidence that the product is fun.
|
|
|
|
## Exit boundary
|
|
|
|
11.3 closes only when all five segments have evidence and the player-facing
|
|
single-region probe works against the authoritative server. 11.2 owner routing
|
|
must also be closed before M11 claims persistent control across regions. Mutation,
|
|
funded budding, complete permadeath, logout/resume and repeated-life playtesting
|
|
remain later gates; the 11.3 probe does not pretend those paths exist.
|