There's been some threads on fetch-on-render vs render-as-you-fetch and what has been possible and not before Suspense (and concurrent features). Some pretty advanced patterns have been possible even before, here are some that I've used. 🧵

First, on the server you've basically had two options, fetch everything before rendering, or render multiple times. (At least if you ignore my experiments with the custom async SSR renderers react-aldrin and react-lightyear, but those were always experiments.)

Huuge news! 🎉 I'm most impressed by the 100% incrementally adoptable part, finding a good solution for that while fully embracing the future can't have been easy. Very excited to read the RFC. 👀

I ❤️ this categorization of architectural choices for server rendered apps(/frameworks), it simplifies a complex space in a practical way and provides a great starting point. While some are uncommon, all these combinations are possible, but it gets even more complex.. 🧵

Traditionally it has been very tricky to refactor between these different approaches and also to mix them within a single app. This is slowly changing, which is a great thing! For this to be possible, third party libraries also needs to support a bunch of different approaches.