@bacon.bsky.social can't find any updates since the SDK 52 release, what's the latest?
06.08.2025 08:44 β π 0 π 0 π¬ 0 π 0@tom.sherman.is
Software Engineer in Norwich. he/him Working on @frontpage.fyi https://tom-sherman.com
@bacon.bsky.social can't find any updates since the SDK 52 release, what's the latest?
06.08.2025 08:44 β π 0 π 0 π¬ 0 π 0What's the latest on React Server Component support within expo?
05.08.2025 14:02 β π 3 π 0 π¬ 1 π 0It's also not well supported with data fetching libs right now. swr supports it well, although a bit clunky. React Query's support is also a bit awkward. Both library's have very little docs on how to build this pattern
05.08.2025 10:57 β π 1 π 0 π¬ 0 π 0It's probably the most advanced because it requires you to prime your client-side data cache with either a promise or the data itself from the server. Meaning you need to understand both the patterns in your article, plus client side fetching
05.08.2025 10:57 β π 1 π 0 π¬ 1 π 0"Server + client" is useful when you want to refresh the UI separately to refreshing your entire route (by revalidating your server components).
05.08.2025 10:57 β π 1 π 0 π¬ 1 π 0Nice! I think this covers the "server only data fetch" pattern well, but I think there are two other patterns for fetching data with RSCs:
1. Server only
2. Server + client
3. Client only
You covered 1, but 2 and 3 are also crucial. Especially 2 which is another new pattern
So I either find a way to use a fake supervisor or find a way to deploy them to separate machines (no thanks). Both require probably more effort than it's worth
02.08.2025 07:08 β π 1 π 0 π¬ 0 π 0Just realised you said "requiring" and "not supporting" pid 1 is tricky. Yeah... It's because the knot server is built as a few separate processes that need to be supervised
02.08.2025 07:08 β π 1 π 0 π¬ 1 π 0I think they have a way around this now with container support community.fly.io/t/using-cont...
I'll probably give this a go, should be less effort than trying to get knot to run without pid 1
It's because they don't actually run docker
fly.io/blog/docker-...
Trying to get the latest @tangled.sh releases to run on @fly.io and it's proving difficult. Lots has changed since I first set it up
The community maintained dockerfile has a bunch of stuff I don't understand, it also requires pid 1 which doesn't work on Fly
It'll be equivalent in terms of the promises but not using Promise.all is probably preferred because it allows react to start rendering each independent subtree as early as possible
31.07.2025 21:43 β π 1 π 0 π¬ 1 π 0Did you know you can pass a promise to children? So you don't need to split into components just to get a loading boundary, just do
<Suspense>
{promise.then(data => <whatever data={data} />)}
</Suspense>
Or just pass Promise<ReactNode> as children π
31.07.2025 12:41 β π 3 π 0 π¬ 1 π 0Did you know you can pass a promise to children? So you don't need to split into components just to get a loading boundary, just do
<Suspense>
{promise.then(data => <whatever data={data} />)}
</Suspense>
Yeah that was my first thought (or a hash of the metadata doc contents) but not sure if that's reliable across implementations. Also a bit concerned it will show as multiple apps in the Auth server UI or something
30.07.2025 11:58 β π 0 π 0 π¬ 0 π 0Thought about a query string as a cache buster but not sure if that's reliable.
30.07.2025 11:41 β π 1 π 0 π¬ 0 π 0I'm running into problems with the PDS caching my client metadata document for atproto oauth
If I add a scope, my previous document is cached for some time and PARs fail because the PAR includes the new scope that the cached document doesn't π©
@matthieu.bsky.team any ideas on a fix?
I want it to work but it's just not easy. Am I holding it wrong?
30.07.2025 08:33 β π 1 π 0 π¬ 0 π 0I just find it so frustrating as a user to pay in crypto. Crypto is difficult to acquire in the UK because of regulations, that makes ad-hoc tx expensive. You could think to solve that by converting large amounts on a regular basis, but then you have huge risk because of holding crypto
30.07.2025 08:33 β π 2 π 0 π¬ 1 π 0com.atproto.uno.reverse
29.07.2025 21:46 β π 6 π 0 π¬ 0 π 0HTTP method seems to be pretty meaningless tbh when the error types are declared in the schema
29.07.2025 19:15 β π 2 π 0 π¬ 1 π 0Pretty sure all bsky APIs do that. XRPC does allow 404 for missing resource though
29.07.2025 19:14 β π 3 π 0 π¬ 1 π 0Gonna need
const backwardsCompatUse = p => semverGreater(React.version, "19.0.0") ? React.use(p) : throw p;
In every library now π
The flow in atproto oauth to request more scopes is simply a case of putting the user through login again with the additional scopes in the metadata doc right? There isn't anything more than that?
Do I need to rotate jwks when doing this?
Yeah we wanna add this, all of the core maintainers are busy with day jobs and life stuff right now tho so work has slowed down quite a bit!
28.07.2025 22:10 β π 1 π 0 π¬ 0 π 0Proposal to add changelogs to the pnpm website & docs
28.07.2025 21:54 β π 2 π 0 π¬ 0 π 0Also, an explanation of what actions to take and a way to know if you're affected by each of the breaking changes would be good. It's really hard to understand as a user right now
28.07.2025 17:23 β π 0 π 0 π¬ 1 π 0