AI summary by TwinMind
STREAM VOD: CHAT WITH CHANCE, JOB MARKETS AND MORE - STREAM VOD 11-4-22
Theo - t3․gg
4h 1m 2s
Guest Introduction & Background
- Chance Strickland from Remix Software (now Shopify)
- Twitter: @chancethedev
- Works on Remix full-stack JavaScript framework
- Team was 7-9 people pre-acquisition
- Building course: "Backend for Front-end Devs" at fronttoback.dev
Remix Acquisition by Shopify
- Shopify acquired Remix Software recently
- Plan: Hydrogen team will rebuild Hydrogen on top of Remix
- Will test Remix at scale with biggest e-commerce brands
- Provides opportunity to validate scalability patterns
Full Stack Development Philosophy
- Remix as "center stack framework" - brings front-end and back-end closer together
- Teaches back-end technology through full-stack app development
- Importance of understanding both sides for performance and user experience
- Front-end developers should understand database requests and back-end concepts
Framework Scaling Discussion
- Pushback: patterns may not scale beyond 20+ person teams
- Counter-argument: scaling problems often bureaucratic, not technical
- Vertical team structure (mini-startups) vs horizontal (back-end/front-end teams)
- Remix increases what one developer can own and maintain
Type Safety Implementation
- Evolution from manual casting to generics to built-in type inference
- Chance initially resistant to TypeScript complexity in libraries
- Pedro Cattori designated as TypeScript specialist on team
- Community contributions (Colin McDowell, Maple) pushed type safety forward
- Loader/action type inference now works within same route file
Development Patterns & Primitives
- Remix designed as small set of tools: request handler + compiler
- Philosophy: provide primitives for building higher-level abstractions
- Resource routes equivalent to Next.js API routes but more flexible
- Can build entire APIs using just Remix routes
- Example: Svelte support built using resource routes
HMR (Hot Module Replacement) Challenges
- Major pain point for developers, especially with WebRTC/websocket applications
- Team aware it's needed but resource/priority constraints
- Comparison to use effect: "95% don't need it, 5% impossible without it"
- Server-side HMR much more complex than client-side
- Intentionally didn't ship buggy version - prefer honest "we don't have it"
Compiler Architecture Decisions
- Built custom compiler instead of using Vite
- Vite didn't exist when Remix started, or lacked SSR support
- Rollup took 40 seconds to rebuild with 2000 routes
- Moved to esbuild for performance
- Compiler kept as implementation detail to allow future changes
- No plugin API yet to avoid locking into specific tooling
Application Lifecycle & Use Cases
- Discussion of session length and framework responsibilities
- Remix optimized for interactive sessions with low data throughput
- E-commerce as golden use case
- Acknowledgment that some use cases (like Ping's real-time AV) need different solutions
- Web standards focus vs framework-specific solutions
Community & Open Source
- Strong Discord community with active contributors
- Examples: Maple's type safety work, trpc bindings, electron integration
- Framework agnostic goals - view/svelte adapters planned
- Community building adapters and extensions
