ELANG
Electronic Language — An open, standardized digital format for defining, preserving, and exchanging human and constructed languages.
ELANG is a portable, deterministic definition and packaging format for
constructed languages, together with its Rust reference tooling. This repository
is the reference implementation: the .elang binary container, the KDL 2.0
package.epf descriptor, the KDL 2.0 .eld Linguistic Definition extension
(v0.1.0 ships the phonemes module only), SHA-256 integrity, canonical
reproducible encoding, and a reader/writer/validator SDK plus a CLI.
Documentation
The public documentation site is built with Zensical from
the Markdown sources in docs/public/ and configured by
zensical.toml. It is published to GitHub Pages by
.github/workflows/docs.yml on pushes to main
that touch the docs, and is served at
https://rockydevworks.github.io/elang-rs/.
The internal normative material (the steering contract, ADRs, requirements, and
the open-artifact tracker) lives under docs/ and .kiro/ and is the
source of truth that the public site is derived from.
Status
Pre-release, tracking v0.1.0. Several normative artifacts (the exact approved
IPA allowlist, its canonical SHA-256, and the META-INF/integrity.kdl grammar)
are open artifacts that must be produced and reviewed before the conformance
suite is frozen. See docs/specification.md and the
open-artifact tracker in
.kiro/specs/elang-v0.1.0/requirements.md.
Version domains
ELANG keeps several independent version domains. A change in one does not
imply a change in another.
| Domain | Value | Notes |
|---|
| Specification release | 0.1.0 | The normative spec this code targets |
| Binary container | 0.1.0 | Header major 0, minor 1 |
EPF (package.epf) | 0.1.0 | Descriptor schema |
| ELD (phoneme inventory) | 0.1.0 | .eld schema |
| Language release | strict MAJOR.MINOR.PATCH | Chosen by each language author |
| Canonical compression profile | elang-canonical-zstd-v1 | Reproducible Zstd settings |
The reader supports binary 0.1 only. Header/format.version's first two
components MUST match the header major/minor; the patch component cannot change
the binary grammar.
Supported scope
A compliant phoneme-only package contains exactly these logical resources:
Header/format.version
mimetype
package.epf
META-INF/integrity.kdl
Linguistics/phonemes.eld # path configurable via the EPF manifest
ELANG does not define lexicons, phonotactics, morphology, syntax, semantics,
prosody, audio, GUI, dependency management, repository ownership or publisher
signatures. A verified package digest proves integrity, not authorship.
Workspace layout
crates/
elang-core/ IDs, errors, limits, validated models
elang-container/ header/index/footer, frame I/O, reader/writer
elang-kdl/ constrained KDL 2.0 parsing and canonical emission
elang-epf/ package.epf schema and validation
elang-eld/ phoneme model, IPA, `when` AST, canonical ELD
elang-repertoire/ pinned Unicode/IPA allowlist and verification
elang-integrity/ integrity manifest and SHA-256 verification
elang-conformance/ fixture runner and interoperability checks
elang-cli/ inspect, verify, validate, pack, unpack, canonicalize
specs/ authoritative schemas and registries
fixtures/ positive, negative, byte-exact, fuzz seeds
docs/ normative spec, ADRs, security and CLI docs
.kiro/ steering contract and the v0.1.0 spec breakdown
Building
cargo build --workspace
cargo test --workspace
cargo clippy --workspace --all-targets
Toolchain and MSRV
The build pins channel = "stable" via rust-toolchain.toml.
The minimum supported Rust version (MSRV) is 1.90, declared as
rust-version in every crate and enforced in CI. Raising the MSRV requires an
ADR under docs/decisions/.
CLI (preview)
The elang binary will expose:
elang inspect <file> stages 1–7 (structure only, no integrity)
elang verify <file> stages 1–11 (all resources hashed)
elang validate <file> stages 1–12 (full ELD semantic validation)
elang pack <dir> -o <file> build a package
elang unpack <file> -o <dir> extract safely (no traversal/symlink escape)
elang canonicalize-eld <in> -o <out>
elang repertoire info show pinned chart/Unicode/digest
inspect is not a full integrity check; only verify and validate hash
every resource. The CLI exits non-zero on any validation failure.
Security and trust model
See SECURITY.md. In short: ELANG defends against malicious
archives (decompression bombs, path traversal, parser abuse) with bounded,
streaming, checked-arithmetic code, and it never treats a hash or URN-like
identifier as proof of publisher identity.
License
Licensed under the Apache License, Version 2.0. See LICENSE.
