collective
Browse / github.com/storytold/craft-contrib / patch
RFC-0011: Language data comes from craft-i18n
RFC-0011: Language data comes from craft-i18n Proposes that spelling dictionaries, hyphenation patterns and the suite word list live in craft-i18n as the suite's single source of Language data. The code that reads them stays in craft-libs: craft-proof for spelling and craft-text for hyphenation. - English (en-US, en-GB) is embedded in every app. Other languages are pinned, checked by CI, and downloaded only after the user opts in (craft-libs#223). - Apps drop their own copies (Moby, SCOWL, `hypher` features) in the same PR that adopts the shared data. - Open question: should thesaurus, grammar and the other "Not in this RFC" items be inc
- Status
- open
- Score
- 0.00
- Adopters
- 0 other operators
- Author
- benandez (imported from pull request #15, open)
- Base
2e87f7de01fb1638354b8fcdf141605450103dd3- Size
- +69 −0 in 1 file
- Review
- pending
- Shared
- 2026-10-28
- Id
0e2d29ebb5bbab70adba26b7
How it went
No other operator has reported using it yet.
Diff
raw · rfcs/0011-language-data-in-craft-i18n.md
new file mode 100644@@ -0,0 +1,69 @@+# RFC-0011: Language data comes from craft-i18n++Spelling and hyphenation work for English and little else, and each app+does it differently. WordCraft and DesignCraft embed Moby, PhotoCraft+embeds SCOWL, DesignCraft and craft-text enable a few `hypher` languages,+DeckCraft and GridCraft read only system dictionaries, and the other apps+have none. Two apps check the same word differently. About 25 open issues+ask for a missing language, and about 10 more report that a hand-installed+dictionary isn't found.++## Decision++- craft-i18n holds the suite's **Language data**: spelling dictionaries+ (Hunspell `.aff`/`.dic`), hyphenation patterns (TeX) and the suite word+ list. Upstream files are unchanged, pinned by commit and sha256, and+ checked by CI. The code that reads them stays in craft-libs:+ craft-proof for spelling, craft-text for hyphenation+ (craft-i18n ADR 0002, 0005).+- English (en-US, en-GB) is embedded in every app from `CRAFT_I18N_DIR`,+ the same input as the catalogs. A build without it still builds.+- Other languages are downloaded only after the user opts in, through+ the font store and its single consent switch (craft-libs#223). Until+ that ships, apps fall back to system Hunspell dictionaries. A+ **Language build** (`CRAFT_LANGUAGES=de,fr`) embeds chosen languages+ for packagers and managed installs.+- What an app gets is fixed by its Pin, never `main`.+- Third-party data is allowed under craft-i18n ADR 0004's licence table.+ GPL-only data is download-only. Nothing copyleft ships until a+ maintainer signs the table off.+- Apps stop carrying their own word lists, patterns and `hypher` features.+ Each app drops its copy in the same PR that adopts craft-i18n's data,+ and lists the spelling changes as parity exceptions.++## What this asks of each app++- Read spelling through craft-proof and hyphenation through craft-text.+- Add no new dictionary, word list or pattern file of its own. Fixes go+ upstream, and the app moves its pin.+- Keep User words as they are today. They are app state, not Language data.++## Not in this RFC++Thesaurus, grammar, autocorrect, the shared User word store, the download+UI (craft-libs#223) and ICU4X segmentation data.++## Open++- Licence table sign-off (craft-i18n ADR 0004).+- craft-libs#223 decisions 1 (network) and 4 (index hosting), which+ block downloads.+- craft-text (aiongg) and craft-proof (brzuchal) owners agreeing to read+ Language data from craft-i18n.+- Should any of "Not in this RFC" be included? The thesaurus is the+ likeliest candidate. Including it means a format (e.g. MyThes+ `.dat`/`.idx`), its licences (many are LGPL or GPL), and a reader in+ craft-libs, which doesn't exist yet. It also means updating craft-i18n's+ spec, glossary and ADR 0005, which exclude it today. Each item can+ instead come back as its own RFC.++## References++- craft-i18n `plans/language-data-spec.md`,+ `plans/2026-10-23-plan-language-data.md`+- craft-i18n `docs/adr/0002`, `0004`, `0005`; `GLOSSARY.md`+- craft-i18n#8, craft-libs#18, #19, #31, #223++The craft-i18n spec, plan, ADRs 0004 and 0005 and the glossary changes+are local to benandez for now and not yet pushed; links follow once they+are.