SPEC 00 — Concord Layering, Semantics, and Conformance

File: specs/00-layering-and-conformance.md
Status: Normative
Version: 0.2

Purpose

Define the Concord “core vs semantics” boundary and provide conformance levels that match how the system is intended to be implemented: core protocol is minimal; identity/permissions/encryption are expressed as ledger data and interpreted by deterministic replay.

This spec exists to prevent accidental “baking semantics into core” while still allowing official, portable semantics.

Design Principle

• The ledger is an append-only event log.
• “Tables” (users, permissions, etc.) are materialized views derived by replay.
• Core protocol validates structure, canonicalization, hashing, and replay ordering only.
• All meaning (identity, permissions, encryption) is introduced by semantic layers that define entry kinds and deterministic reducers.

Layers

Layer L1 — Core Protocol

Responsibilities:

• LedgerContainer / Commit / Entry shapes
• Canonicalization + hash addressing
• Commit graph rules
• Deterministic replay ordering

Non-responsibilities:

• Signature verification
• Identity resolution
• Key rotation semantics
• Permissions / authorization
• Encryption

Core treats Entry.kind as opaque.

Layer L2 — Semantic Tables (Identity, Permissions, Encryption)

Responsibilities:

• Define entry kinds that represent rows/updates
• Replay reducers that materialize state maps
• Validation rules that are deterministic from the ledger state
• Optional signature verification and policy enforcement

Semantic layers may be shipped as “official plugins” but remain separate from core.

Conformance Levels

Normative Statement on Rotation

“Key rotation” is a semantic concept expressed via normal entries and replay rules. It is not a core protocol primitive.

• Identity “rotation” is achieved via successive identity entries.