Technical roadmap

Technical roadmap

The roadmap advances from observable facts to controlled ownership, then to handoff and recovery.

Completed foundations

0.0.1 Genesis

Monorepo, licensing, architecture decisions, pinned toolchains, generated contracts, and hosted quality gates.

0.0.2 Protocol Laboratory

Versioned Rust and Java control path with worker registration, heartbeat sequencing, session generations, and stale-session rejection.

0.0.3 Observer

Read-only structural and bounded deep inspection for offline Java Anvil worlds with independent fixture evidence and two-pass fail-closed consistency.

Near-term sequence

0.1.0 Single Worker

Connect Velocity, Sling, one ElahFolia worker, and elahd. Establish a gameplay baseline without sharding.

0.2.0 Static Cells

Run two workers against a pregenerated world with fixed cell ownership. Prevent any worker from loading an authoritative cell it does not own.

0.3.0 Sling Handoff

Transfer a player across a fixed worker boundary without a client reconnect. Use transaction identities, state versions, timeouts, rollback, retries, hashes, and traces.

0.4.0 Ownership Epochs

Attach monotonically increasing epochs to authoritative cell operations and reject stale writers.

0.5.0 Crash-safe Transfers

Inject source, target, proxy, controller, and network failures. Prove no item duplication, impossible concurrent ownership, or unexplained state loss.

Later work

1.x: Safe static multi-machine Minecraft

The 1.x train expands from more workers and drain operations through entity handoff, boundary ghost state, storage, generation, capacity evidence, rolling upgrades, operator tooling, and control-plane high availability.

Elah 1.0 is the first production-supported target: one logical deployment across several workers using static spatial ownership. It does not include dynamic placement or elastic workers.

2.0 through 2.3: Dynamic Minecraft placement

Shepherd gains live cell migration, placement recommendations, optional automatic placement, cell split and merge, and operator placement constraints.

Elah must prove that it can safely change ownership before it may create or remove infrastructure capacity.

2.4 through 2.9: Runtime Providers and elastic workers

Release	Direction
2.4	Runtime Provider contract and conformance laboratory.
2.5	Native elastic workers.
2.6	Optional Docker worker lifecycle.
2.7	COLD, WARM, and HOT pools with GUARANTEED, BURST, and PREEMPTIBLE capacity.
2.8	Hard resource envelopes and authenticated External Provider integrations.
2.9	Qualification of many isolated clusters sharing one heterogeneous infrastructure pool.

2.10 through 2.14: Advanced single-region operation

Predictive prefetch, regenerative storage, advanced Stone, boundary simulation, and published scale qualification follow provider safety.

3.x and 4.0

The 3.x track adds geographic operation, replicated storage, disaster recovery, global directories, cross-region plugin primitives, and mature security. Elah 4.0 represents the complete original vision rather than the end of development.

Dynamic repartitioning, automatic scaling, multi-region operation, custom clients, GPU generation, and a Kubernetes requirement are deliberately not early work.

The sequence is directional, not a promise of dates.

Read the Runtime Provider design.