Internationalization with Project Fluent

rfcs/fluent-internationalization.md

@@ -0,0 +1,81 @@
+# Feature Name: `fluent-internationalization`
+
+## Summary
+
+Internationalization is the practice of adapting your app's user interface to support multiple languages.
+This RFC proposes how we can create a first-class integration with the [`fluent-rs`](https://github.com/projectfluent/fluent-rs) crate.
+
+## Motivation
+
+Internationalization is an important feature for serious games and apps, allowing them to expand their audience beyond a single language.
+The technical challenges are well-understood, and not insurmountable, but retrofitting existing designs to accommodate them is notoriously painful.
+
+Even for unilingual games, a high-quality internationalization language provides excellent support for dynamically changing text:
+interpolating names, adapting nearby words to the correct plurality and gender, date and time formatting and so on.
+This is very useful for building polished UIs in games, where the details of what a widget or dialog box should display often vary dramatically based on player actions.
+
+## Guide-level explanation
+
+TODO: Write.
+
+Language Resource.
+
+Labels for text and audio.
+
+Assets should use real text if possible.
+
+## Reference-level explanation
+
+TODO: Write.
+
+## Drawbacks
+
+1. The data model for unilingual text is necessarily more complex.
+2. Our dependency tree will grow.
+3. We're forced to confront challenges around variable text length, left-to-right text, non-Latin glyphs etc. sooner.
+
+## Rationale and alternatives
+
+### Why does Bevy itself need internationalization?
+
+1. By establishing a single standard, we can ensure that ecosystem crates (for e.g. dialog, widgets) all interoperate correctly.
+2. Bevy should have text-enabled widgets in `bevy_ui`. If we don't support localization on those, they will quickly be obsoleted by third-party equivalents.
+3. Bevy itself is responsible for solving many of the ancillary problems induced by variable text contents, such as automatically rescaling UIs.
+4. By enabling internationalization by default, games can seamlessly mature, rather than having to rip out all related prototyping work.
+5. Existing internationalization solutions in other game engines are typically bolted-on and second-class. Excellent integration with a high-quality internationalization solution offers a serious competitive advantage for multilingual indie developers and serious commercial game developers considering Bevy.
+
+### Why Fluent?
+
+TODO: Write.
+
+## Prior art
+
+An existing Bevy integration with `fluent-rs`, [`bevy_fluent`](https://crates.io/crates/bevy_fluent) shows that basic integration is quite simple.
+
+[Project Fluent](https://projectfluent.org/) has a great overview explaining the limitations of simpler approaches to localization.
+
+### Internationalization in other game engines
+
+#### Unity
+
+TODO: Research.
+
+https://medium.com/i18n-and-l10n-resources-for-developers/localizing-unity-games-with-the-official-localization-package-5e2f59779d45
+
+#### Unreal
+
+TODO: Research.
+
+https://docs.unrealengine.com/en-US/ProductionPipelines/Localization/Overview/index.html
+
+#### Godot
+
+TODO: Research.
+
+https://docs.godotengine.org/en/stable/tutorials/i18n/internationalizing_games.html
+
+## Unresolved questions
+
+1. How do we ensure that the "simple case" of code-only static text in a single language is still easy?
+2. How do we ensure that translators have appropriate access to context?
+3. How do we generalize over different asset types?