Architecture And Internals

This section explains how Nervix is built and how it behaves internally.

Use it for:

• control-plane and data-plane structure
• execution primitives: backend selection, native and browser capabilities, permissions, and verification boundaries
• data-plane concurrency: publication, ownership, ordering, production-owner models, and diagnostic locks
• execution plans: the committed schedule, typed revisions, local binding, publication, and recovery
• typed failure ownership, the report model, fixed outcome boundaries, sensitivity, recovery, enforcement, and public diagnostics
• typed absence, semantic states, validation boundaries, state identity, and boundary encodings
• connector crate ownership, the shared source and sink contract, and host execution
• HTTP emitter request preparation, delivery, lifecycle, and qualification
• client ingestors and emitters: native batches, application ACKs, credits, forwarding, restoration, buffered ALTER and paced workflows
• domain-clock mapping, authority, lifecycle, progress, and execution-time semantics
• cluster interconnect security, traffic isolation, and delivery semantics
• host name resolution: the node’s asynchronous DNS resolver, its configuration, cache and bounds, the client-library hooks and every call site, residual driver and browser resolution, deployment limits, and the evidence behind them
• client-to-node communication: the FlatBuffers session protocol, request correlation and dispositions, exact recovery, redirect and reconnection, Row subscriptions, uploads, and the shared client binding
• transaction impact planning, scoped quiescence, inspection, and retained outcomes
• consensus durability, replication pacing, log retention, and snapshot recovery
• resource versions, pinned bindings, LATEST resolution, rebinding, and the HTTPS listener refresh
• WASM guest state: checkpoints, generations, resets, rejected-state recovery, and replay
• shutdown phases, drain guarantees, and crash recovery
• integration-test lifecycle: harness deadlines, node startup, teardown, and the suite watchdog
• deterministic interconnect simulation: the Turmoil build mode, fault model, seeded replay, failure records, scenario matrix, and CI budget
• property testing and fuzzing: registered Bolero targets, current-domain corpora, ordinary regressions, sanitizer libFuzzer runs and exact failure replay
• the expression VM: compilation, columnar execution, kernels and SIMD, function families, window aggregates and sketches, and adding a function
• SIMD kernels: typed buffer ownership, dispatch and build targets, exact results, caller catalog, qualification, and measured evidence
• runtime semantics that are easier to understand from the implementation side
• relay/state internals

If you are looking for NSPL usage, setup, or feature-level guidance, start in the Manual.