AI summary by TwinMind
The First Real Webpack Alternative (Written in Rust!)
Theo - t3․gg
19m 28s
RS Pack 1.0 Alpha Release Overview
- RS Pack 1.0 Alpha now available on npm
- 1-2 months testing period planned before stable release
- Stable version 1.0 expected August release
- Rust rewrite of webpack with major features/APIs implemented
- Alternative to Turbo Pack (Vercel's webpack rewrite project)
Key New Features & Improvements
- Module concatenation enabled by default in production builds
- Bundles npm modules into single vendor module
- Reduces parsing/execution overhead in browser
- 4-10% code size reduction
- Lightning CSS support added
- CSS bundler built in Rust (extremely fast)
- Implements modern CSS features, Sass/Less compatibility
- Tailwind V4 will be built entirely around Lightning CSS
- Lean core architecture
- Removed non-core features (SWC for emotion, styled-components, relay)
- Features still available via experimental plugins
- Core focused on common bundler features
Performance Comparisons & Benchmarks
- Bite Dance internal projects: 5-10x improvement over webpack
- Build time examples:
- Webpack startup: 8 seconds, HMR: 300ms
- RS Pack startup: ~200ms, HMR: varies
- Production builds: 7 seconds → 600 milliseconds
- Real-world impact: Production build times reduced from 10-30 minutes to much faster
- Cold start performance: Previously several minutes, now significantly faster
Migration & Compatibility
- Webpack compatibility: Compatible with webpack config schema & loader architecture
- Legacy loader support: Can use existing Babel, Less, Sass, Vue loaders
- Migration guides available: Detailed step-by-step from Create React App
- Flexible configuration: Accommodates real-world project requirements
Technical Architecture & Background
- Built in Rust for performance benefits
- Highly parallelized architecture
- Built-in HMR optimization (webpack had no real HMR concept)
- Memory-level caching currently (stronger caching planned)
- Created by Bite Dance (TikTok) to solve massive codebase performance issues
- Zach's involvement: Moved from Lululemon to Bite Dance to focus on large-scale bundling problems
Ecosystem Comparison Analysis
- vs Vite: Vite still relies on Rollup (JavaScript-based), slower than Rust solutions
- vs ESBuild: ESBuild lacks complete feature set (no HMR, limited incremental compilation)
- vs Turbo Pack: Turbo Pack redesigned architecture but steeper migration cost, not production-ready
- vs Farm: Another Rust-based bundler mentioned, potentially faster than RS Pack
- Bundler fatigue: Multiple Rust rewrites happening simultaneously (Rolldown for Rollup, etc.)
Development Experience & Tooling
- RS Build: RS Pack-powered build tool (like Vite to ESBuild/Rollup relationship)
- Biome integration: Rust-based ESLint/Prettier replacement recommended
- Binary size impact: Rust tools ship 20MB+ binaries vs KB JavaScript packages
- PNPM ecosystem adoption: Community standardizing on PNPM over NPM/Yarn
- Instant dev server: 0.02 second reload times demonstrated
Future Goals & Roadmap
- Community collaboration: Not just Bite Dance internal tool
- Enhanced plugin capabilities: Better than webpack loaders
- Continued performance improvements
- Compatibility-first approach: Focus on existing project migration
