Follow BCP 47 for language codes: tags in the code and the tests, folder names only where the browser demands them #12

New issue

Closed

opened 2026-10-01 06:39:30 +00:00 by tiagoagueda · 0 comments

tiagoagueda commented

2026-10-01 06:39:30 +00:00

Owner

The requirement: all language codes must follow BCP 47. A BCP 47 tag (RFC 5646) is made of subtags from the IANA registry joined by hyphens, and §2.1.1 gives the form to write one in: the language in lower case, a script in title case, a region in upper case (pt-BR, en-GB, sr-Cyrl). The rule for the whole project is postulo/postulo#337; this is what it means for the extension, whose source this repository holds for both browsers.

What exists today

Already canonical. Each catalogue's languageTag is a tag in the right form (pt-BR, pt-PT, nb, fr). It is what the pages put in <html lang> (src/lib/i18n.js:23), what dates are formatted with on the list page (Intl.DateTimeFormat(t("languageTag")), src/captures.js:177), and what the downloadable log's header says (src/options.js:48).
Forced by the browser. The folders under src/_locales/ are named the way the WebExtension format demands, with an underscore: pt_BR, pt_PT. default_locale in the manifest names a folder. These stay.

Where it does not follow

The test holds folder names, not tags. OFFERED in test/i18n.test.mjs lists "pt_BR" and "pt_PT", and the check that a catalogue names its own language is folder.replace("_", "-"): it swaps one character and checks neither case nor validity.
One date ignores the tag. The popup's "captured before" date is toLocaleDateString() with no locale (src/popup.js:204), so it follows the browser, not the catalogue in use. postulo/postulo-chromium#11 names the same line.
A comment writes the core's codes as locale names (fr_FR, en_GB, test/i18n.test.mjs:11).
Two tags for one catalogue. The extension says fr and en where the core says fr-fr and en-gb (fr-FR and en-GB once postulo/postulo#337 lands). Both are valid; the mapping is written nowhere but that comment.

What the extension should do

The list of languages is a list of BCP 47 tags in canonical form, and a folder's name is derived from its tag (hyphen to underscore) in one place. The test asserts each languageTag is canonical, for instance Intl.getCanonicalLocales(tag)[0] === tag, and that the folder is the one derived from it.
Every date and every lang attribute uses the tag of the catalogue in use.
The mapping from the core's tags to the extension's is written down once, with its rule: the region is dropped where the core has one variety of a language, kept where it has two. postulo/postulo-chromium#10 proposes the script that writes the list from the core; it is the place for it.
Whatever a person chooses or the extension sends is a tag, never a folder name: the stored display language of postulo/postulo-chromium#11, and an Accept-Language header if that issue adds one.
The README's Languages section states the rule.

Decisions

Serbian. The catalogue is Cyrillic and its tag is sr. If postulo/postulo#337 settles on sr-Cyrl, languageTag can say so while the folder stays sr, the name browsers match. Recommended: follow the core.
fr or fr-FR. Recommended: keep fr and en: a browser set to any French matches the fr folder, and the tag should name the catalogue actually served.

postulo-firefox needs no change: it holds no language code and takes this through its pin. No new on-screen sentence is expected; if one appears, test/i18n.test.mjs requires it in all 40 _locales at once.

Done when

npm test fails on a languageTag that is not a canonical BCP 47 tag, and on a folder whose name is not derived from its tag.
The popup's dates follow the catalogue in use.
The README says what a language code is here and where the underscore form is the browser's.

Requested on 2026-10-01: all language codes must follow BCP 47.

The requirement: **all language codes must follow BCP 47.** A BCP 47 tag (RFC 5646) is made of subtags from the IANA registry joined by hyphens, and §2.1.1 gives the form to write one in: the language in lower case, a script in title case, a region in upper case (`pt-BR`, `en-GB`, `sr-Cyrl`). The rule for the whole project is postulo/postulo#337; this is what it means for the extension, whose source this repository holds for both browsers. ## What exists today - **Already canonical.** Each catalogue's `languageTag` is a tag in the right form (`pt-BR`, `pt-PT`, `nb`, `fr`). It is what the pages put in `<html lang>` (`src/lib/i18n.js:23`), what dates are formatted with on the list page (`Intl.DateTimeFormat(t("languageTag"))`, `src/captures.js:177`), and what the downloadable log's header says (`src/options.js:48`). - **Forced by the browser.** The folders under `src/_locales/` are named the way the WebExtension format demands, with an underscore: `pt_BR`, `pt_PT`. `default_locale` in the manifest names a folder. These stay. ## Where it does not follow 1. **The test holds folder names, not tags.** `OFFERED` in `test/i18n.test.mjs` lists `"pt_BR"` and `"pt_PT"`, and the check that a catalogue names its own language is `folder.replace("_", "-")`: it swaps one character and checks neither case nor validity. 2. **One date ignores the tag.** The popup's "captured before" date is `toLocaleDateString()` with no locale (`src/popup.js:204`), so it follows the browser, not the catalogue in use. postulo/postulo-chromium#11 names the same line. 3. **A comment writes the core's codes as locale names** (`fr_FR`, `en_GB`, `test/i18n.test.mjs:11`). 4. **Two tags for one catalogue.** The extension says `fr` and `en` where the core says `fr-fr` and `en-gb` (`fr-FR` and `en-GB` once postulo/postulo#337 lands). Both are valid; the mapping is written nowhere but that comment. ## What the extension should do - The list of languages is a list of BCP 47 tags in canonical form, and a folder's name is derived from its tag (hyphen to underscore) in one place. The test asserts each `languageTag` is canonical, for instance `Intl.getCanonicalLocales(tag)[0] === tag`, and that the folder is the one derived from it. - Every date and every `lang` attribute uses the tag of the catalogue in use. - The mapping from the core's tags to the extension's is written down once, with its rule: the region is dropped where the core has one variety of a language, kept where it has two. postulo/postulo-chromium#10 proposes the script that writes the list from the core; it is the place for it. - Whatever a person chooses or the extension sends is a tag, never a folder name: the stored display language of postulo/postulo-chromium#11, and an `Accept-Language` header if that issue adds one. - The README's *Languages* section states the rule. ## Decisions - **Serbian.** The catalogue is Cyrillic and its tag is `sr`. If postulo/postulo#337 settles on `sr-Cyrl`, `languageTag` can say so while the folder stays `sr`, the name browsers match. Recommended: follow the core. - **`fr` or `fr-FR`.** Recommended: keep `fr` and `en`: a browser set to any French matches the `fr` folder, and the tag should name the catalogue actually served. postulo-firefox needs no change: it holds no language code and takes this through its pin. No new on-screen sentence is expected; if one appears, `test/i18n.test.mjs` requires it in all 40 `_locales` at once. ## Done when - `npm test` fails on a `languageTag` that is not a canonical BCP 47 tag, and on a folder whose name is not derived from its tag. - The popup's dates follow the catalogue in use. - The README says what a language code is here and where the underscore form is the browser's. Requested on 2026-10-01: all language codes must follow BCP 47.

tiagoagueda added the

enhancement

label

2026-10-01 06:39:30 +00:00

tiagoagueda referenced this issue from Postulo/postulo

2026-10-01 06:49:04 +00:00

Write every language code as a BCP 47 tag in its canonical form: pt-BR, not pt-br or pt_BR #337

tiagoagueda referenced this issue from a commit

2026-10-01 21:18:23 +00:00

Name a language by its BCP 47 tag, and its folder only where the browser asks

tiagoagueda closed this issue

2026-10-01 21:18:23 +00:00