In order to remove CLD2, we'll need to migrate to the new fastText language detector, which the translations infrastructure will rely upon. This will involve a download for the first time it is run, since we don't ship with the wasm blob or language models. This could cause a behavior change, or for the language detection to fail if the download fails. I'm going to block on Bug 1813161 now since this touches the language identification code. Here is a [searchfox search](https://searchfox.org/mozilla-central/search?q=LanguageDetector.jsm&path=&case=false®exp=false) of consumers. * The extension API uses `browser.i18n.detectLanguage`. * Reader mode uses it to detect the article language. :zombie, :mixedpuppy As module owners of the web extensions, do you have any issues with me moving to use our new language detection library? It's using `fastText`, and should provide more accurate results. The trade off is that it requires a download upon the first use. I'm not seeing code owners around a decision for Reader Mode, so I'll rely on my reviewer. The downloads for fastText are a 875k wasm, and 938k language model.
Bug 1817084 Comment 4 Edit History
Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.
In order to remove CLD2, we'll need to migrate to the new fastText language detector, which the translations infrastructure will rely upon. This will involve a download for the first time it is run, since we don't ship with the wasm blob or language models. This could cause a behavior change, or for the language detection to fail if the download fails. I'm going to block on Bug 1818603 now since this touches the language identification code. Here is a [searchfox search](https://searchfox.org/mozilla-central/search?q=LanguageDetector.jsm&path=&case=false®exp=false) of consumers. * The extension API uses `browser.i18n.detectLanguage`. * Reader mode uses it to detect the article language. :zombie, :mixedpuppy As module owners of the web extensions, do you have any issues with me moving to use our new language detection library? It's using `fastText`, and should provide more accurate results. The trade off is that it requires a download upon the first use. I'm not seeing code owners around a decision for Reader Mode, so I'll rely on my reviewer. The downloads for fastText are a 875k wasm, and 938k language model.