6. Dual licence: MIT code, CC BY catalogue

Amended by ADR 0024. The code moved from MIT to Apache-2.0 before the first release. The catalogue half of this decision, and the data rule in it, still stand.

Context

The catalogue is a contribution in its own right. The field is littered with one-off toys and unlicensed word lists, and a clean, sourced catalogue of several hundred procedures is useful independently of the code that checks them.

Decision

Code is MIT. The catalogue is CC BY 4.0 with per-entry source attribution. Language data must be audited separately before it lands: Wiktionary-derived data is CC BY-SA and share-alike attaches; pyphen is GPL/LGPL/MPL tri-licensed. Anything copyleft stays behind an extra so the core remains permissive.

Consequences

The catalogue can be cited and reused by people who will never install the package. Core stays permissively licensed no matter what a [de] or [fr] extra later carries. The obligation this creates is real: no lexicon may be added to core, ever, without first recording its licence in an ADR of its own.