Simulation layer policy (summary)

Policies that keep the org coherent

dotnetgovernancenovolis

Operational summary for agents and maintainers. Full boundaries: library-boundaries.md.

Stack (closed)

Math → Physics → Simulation

Math facets: Novolis.Math.Arrays, Novolis.Math.Geometry, future Novolis.Math.Topology.

Simulation facets include Novolis.Simulation.Racing (headless racing sim; ML-agnostic). Neural evolution on racing lives in apps, not in novolis-machinelearning.

Repo	In stack?	Role
novolis-math	Yes	Numbers, geometry, topology — never time
novolis-physics	Yes	Physical evolution with `dt`
novolis-simulation	Yes	Worlds, systems, clocks, all cameras, replay
novolis-raylib	No	Separate graphics host — no dep on Simulation, Simulation does not dep on Raylib
Apps	Compose	Wire stack + Raylib (and product rules) at app layer

Dependency rule

physics → math
simulation → math, physics

Forbidden: Novolis.Raylib.* ↔ Novolis.Simulation.* (either direction). Same for dogfooding/SCR inside platform Simulation packages.

Type rules

BCL first, always: System.Numerics.Vector3, Quaternion, Matrix4x4 — no Vector3d / Novolis.Physics.Numerics duplicates.
No Vector2 in stack; planar XZ = Vector3 with Y = 0.

Time rule

Need	Home
No time (static spatial/math)	Math
`dt`, integration, forces, motion	Physics
Tick order, scenario clock, replay	Simulation

Cameras

All platform cameras → Simulation.

Apps: product-specific feel (run bobbing, recoil); bridge Simulation observers to Raylib Camera3D — not in either library.

SCR vs platform

StarConflictsRevolt.Server.Simulation = product. Novolis.Simulation.* = platform.

HexGame-aligned loops

Orchestration for HexGame-shaped ticks (commands in → advance authoritative state → snapshots/effects out) lives here and in app application cores — not in Physics. See hexgame-authoritative-core.md.