fix(handler): return promise of stream with allReady

[dai-shi]

Fixes #1082
Reverts #1078

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Updated (UTC)
waku	✅ Ready (Inspect)	Visit Preview	Dec 22, 2024 1:59am

[codesandbox-ci]

This pull request is automatically built and testable in CodeSandbox.

To see build info of the built libraries, click here or the icon next to each commit SHA.

[dai-shi]

I think we can revert #1078 with this change. cc @rmarscher

[rmarscher]

I think we can revert #1078 with this change. cc @rmarscher

allReady waits for all of the suspense boundaries to render. #1078 just waits for the first chunk to start sending. They are not the same. I think it's best to keep the waitForFirstByte to allow more flexibility with running server-side code before the first suspense boundary.

[dai-shi]

I think we can revert #1078 with this change. cc @rmarscher

allReady waits for all of the suspense boundaries to render. #1078 just waits for the first chunk to start sending. They are not the same. I think it's best to keep the waitForFirstByte to allow more flexibility with running server-side code before the first suspense boundary.

I understand it's not the same, but waitForFirstChunk is very limited either.
Isn't it correct that handler now behaves like how rsc/ssr did previously?

---

If waitForFirstChunk is truly what peopole want, I think it can be implemented with a separate middleware.

[rmarscher]

Isn't it correct that handler now behaves like how rsc/ssr did previously?

Ah, ok. Yes. I just tested with v0.21.10. This is great. The handleRequest promise now waits on the
initial entry layout and page render functions. If I add a suspense boundary to a layout, it will send the outer shell when it is ready. Perfect.