Questions About SSE and Fetch Behavior in Datastar

[surfskidude]

Datastar looks very interesting and may be a great fit for real-time device management. However, I have a few questions about the SSE concept and client-server interactions since I need to implement a new plugin for our server:

• SSE Connections: Does the Datastar server-side open a new persistent SSE connection for each Fetch call initiated from the client? If so, what happens to any existing SSE connections? Are they closed, or do they remain active until explicitly terminated?
• Client-to-Server Fetch Requests: How are client-to-server Fetch requests handled? The examples include various URLs. Should I interpret this as a REST API design, even though the server responds with SSE streams?
• Would you consider creating a plugin architecture for client-server interaction? Using WebSockets would be preferred.

[bencroker]

Does the Datastar server-side open a new persistent SSE connection for each Fetch call initiated from the client? If so, what happens to any existing SSE connections? Are they closed, or do they remain active until explicitly terminated?

The Datastar library is frontend only. Each fetch results in an SSE connection. Whether the connection is persistent or not depends entirely on the backend implementation.

Client-to-Server Fetch Requests: How are client-to-server Fetch requests handled? The examples include various URLs. Should I interpret this as a REST API design, even though the server responds with SSE streams?

They are fetch requests to the URL of your choosing. The backend API is entirely up to you. All HTTP verbs are available.

Would you consider creating a plugin architecture for client-server interaction? Using WebSockets would be preferred.

SSE has many advantages over WebSockets. You’re welcome to create your own custom Datastar plugin, but WebSockets is unlikely to ever be a first-party thing. Please read Streaming Backend for more context.

[surfskidude]

Thank you for your response to my questions. However, I still have some concerns regarding best practices and guidelines for effective integration.

Since each fetch request results in a new SSE connection, could you provide general guidelines or recommended best practices for managing these connections on the backend, assuming you want a full duplex, always-connected channel between client and server?

In general, we have worked with devices such as the one illustrated in the following image, where the reactive state of the entire application is replicated between multiple views (browsers) and the shared resource on the server. For this reason, WebSockets has been the only viable option without delving into the implementation details.

[bencroker]

Please see the TODOS demo at the bottom of the page at https://data-star.dev/. If you open it in multiple browser tabs, you’ll see that changes are streamed to all clients via SSE (using a similar flow to that depicted in your illustration).