Two licences, one repository

Code and corpus live in the same repository and are licensed apart, because they are different kinds of thing and a consumer usually wants one of them.

what	licence
code — the tooling, and everything outside regimes/	Apache-2.0
corpus data — every regime's canon/, atoms/, graph/, releases/	CC-BY-4.0

Why CC BY 4.0 and not ODbL

A corpus of law is a database, and in some jurisdictions a database attracts a sui generis right on top of copyright. CC BY 4.0 covers both in one instrument and asks only for attribution.

The Open Database License would have asked for more: share-alike on derived databases. That single clause would stop a bank embedding the corpus in a proprietary compliance system — which is precisely the use we are building for. Public-law data that cannot be used inside the systems that have to comply with it is not open in any sense that matters.

The split is machine-readable in three places

A licence stated only in prose gets guessed at by the next packager down the line. This one is stated where machines read:

1. Every source file in the tooling carries SPDX-License-Identifier: Apache-2.0 on its first line, and a check gates it.
2. Every built release names both identifiers in its release-meta.yaml — code: Apache-2.0, data: CC-BY-4.0.
3. The PyPI wheel's License: field, the npm package.json, and both packages' manifest.json are rendered from that block, so a downstream packager reads the split instead of inferring it.

Neither licence relicenses the law

This is the part that is easy to get wrong. OpenRegs does not own the legislation it mirrors, and no grant it makes can enlarge what a regulator's publication service granted. The project's NOTICE states the terms each service publishes under — EUR-Lex, legislation.gov.uk, the eCFR and the Federal Register — and a regime's SOURCES.lock names the exact document, digest and retrieval instant behind every canonical text. What is licensed here is the project's own contribution: the normalisation, the consolidation, the atoms and the graph.

What that means for the commercial line

Two commitments follow from the split, and they are the reason it is written down this way rather than left to good intentions:

• Public-law corpus data is always open.
• Anything required to self-host stays open source.

The commercial line is drawn at operations — hosting, freshness, private regimes, attestation — not at the data and not at the engine. A licence you can verify is worth more than a promise you cannot.