I find it hard to believe this is about money. Maybe he knows something I don't, but this seems like a bad business decision to me.
IMO, this is more ideological.
I find it hard to believe this is about money. Maybe he knows something I don't, but this seems like a bad business decision to me.
IMO, this is more ideological.
It's going to be a lot more difficult to get the motivation to help people with Next or to defend it. I can't believe he would do this.
I feel bad for Vercel employee's.
The inconsistent data in Next hasn't really been a problem for me yet, but it would be nice to have.
The prefetching in Next uses the Convex usePreloadedQuery hook and the prefetching in Start uses react query hooks like useSuspenseQuery, so I'm not sure how difficult it would be to implement this in Next with Convex hooks.
Yeah, prefetching Convex queries in TS Start loaders came to mind while reading this article. I've been working with Start and Convex lately and loving it.
I think prefetching Convex queries in Next RSCs still has this issue with inconsistent data. Do you plan on using a timestamp for this as well?
I hope you feel a little better talking about these things on bluesky where it's not so bad. I know itβs tough to filter out the negativity, but a lot of people really value what you have to say. I definitely do.
I get that you've got the team to think about too, though.
I think a lot of people assume RSCs generate HTML and it causes this confusion. I see it all the time on reddit. That's why they think of it like SSG/SSR.
What do you mean by #4?
Me and my partner have the flu! She's a teacher so we can't avoid it. It's pretty awful.
I don't care about the part of your post that makes assumptions about what the react team actually believes instead of what they say in the docs.
You don't have to use RR7 or RSCs. You can build however you like, with a framework or create your own framework. Up to you.
react-router will support RSCs soon. You will be able to use them in a SPA without a server.
Interesting. I wouldn't have expected that. I assume you can disable it right?
So you believe they shouldn't recommend using a framework until they have their own framework?
Also, they are recommending react-router as a "framework". They are even recommending to use it to build a SPA.
I don't think they are saying a SPA is an unusual constraint. They are recommending react-router to build SPAs. It seems they think of react-router as a framework since it allows you to opt-in to framework features.
I think react-router is CSR by default and it's recommended in the docs as a framework.
They are recommending react-router as a framework. Most react apps will need a router.
react-router allows you to opt-in to framework features, so I assume they consider it a "framework".
I don't think React was ever MVC. It was inspired by XHP.
This confirms it, Devin really is going to take my job.
Are you arguing it's possible to use RSCs and increase bundle size more than the same app without RSCs?
I imagine it's possible, but do you have an example?
Also, I don't think that's what people mean when they say RSCs don't send JS to the client. The JS in the component doesn't go to client.
With RSCs, you are just executing some of the components on the server instead of all on the client. But, RSCs also get sent to the client twice. Once as HTML from SSR and again in the .rsc payload. React needs the .rsc data to know where hydration should occur.
Yes, I often see this criticism that a "double fetch" happens when using RSCs in Next. But this is how things worked before RSCs. In pages router, the client gets HTML generated from the components markup using SSR and components are also hydrated / rendered on the client, so they are in the bundle.
The .rsc payload contains no JS. It contains the element tree for RSCs, "holes" for client components, serialized props, and URLs to scripts for client components.
That doesn't mean no JS is being sent to the client, but it's definitely less JS since RSCs don't execute on the client.
As someone that grew up poor in rural Ohio, it's difficult for me to say "yes". I couldn't imagine paying that much money if I was just starting out. So, it depends.
It's a great course. Maybe you can get an employer to pay for it.
RSCs are rendered on the server, so JS isn't sent to the client for those components, reducing bundle size. The .rsc payload is serialized into a JSON-like format and included in the HTML payload.
Sure, .rsc payload has a cost, but is this really an echo chamber issue?
I don't think much is going to change in a year when it comes to RSCs.
My guess is that Next will continue to be a popular framework and more Next projects will be using App Router. Also, other frameworks will adopt RSCs and make it available even for SPAs. They aren't going anywhere.
I didn't realize the react native app I am typing this on is slow and terrible.
I see criticism about React all the time and it has always been controversial. I honestly don't care to see more of it. There is always Primeagen for that.
With that said, if there is a new criticism that doesn't seem to have an answer yet, I say share it.
Josh's CSS course is one of the best courses I have ever taken.
- followed you on twitter
- I just saw you in replies
I occasionally enjoy it, as I like talking about stuff and helping others. The next subreddit is generally a good community. There are many new developers asking questions and looking for assistance, but I don't mind that. The react subreddit is usually insufferable though. Both require patience π