WebSockets and other realtime communication

[canadaduane]

Is Spin suitable to realtime communication? I've been poking around the edges to see if it would be able to support WebSockets, for example, but there is only a very little information available, mainly this PR:

• SIP 016 - Inbound WebSocket Support #2212

Alternatives that may work for realtime communication (but would be inferior to WebSockets for my case):

• SSE (Server Sent Events)
• HTTP long-polling (maybe)

Can anyone weigh in here on behalf of someone who is completely new to Spin? Thanks!

[itowlson]

Hi @canadaduane, thanks for bringing this up.

Unfortunately, we do not have WebSocket support at the moment. (@dicej did that SIP end up going anywhere? Should we pick it up after the WASI P3 work is released?) I believe it's possible to put a WebSocket bridge in front of Spin, but I'm not familiar with the details (again I think @dicej is the one who's looked into this).

The upcoming Rust SDK has a sample for server sent events (https://github.com/spinframework/spin-rust-sdk/tree/main/examples/server-sent-events). This will work with Spin 4, which should be out next week,or the canary release that you can get from GitHub (https://github.com/spinframework/spin/releases/tag/canary). It should be reasonably easy to port to other languages once we release Spin 4 libraries for those. HTTP polling should also work in Spin 4/canary, and we can support you with that.

(Both of these can be made to work in Spin 3, but it's more of a faff. Spin 4 async and concurrency greatly simplify this sort of thing!)

However, both SSE and long polling mean you are dealing with long-lived components. These work fine in the CLI and SpinKube, but there are other hosts that limit connection times. Long term we'd like to find a WebSocket solution that plays nicely with Spin's naturally stateless idiom, but we're not there yet...

Hope this helps - sorry not to have a more positive answer for you, but I'm happy to discuss possible approaches further, or try to answer any questions!

[canadaduane]

Interesting, thank you for this substantive reply! I'll have to think over whether SSE on Spin 4 would be viable for various projects.

For context, I'm the author a small library called Kyneta which is sort of a capstone project based on earlier work in loro-extended. I don't expect this to mean anything in particular to you, except that there is a very specific pattern that I'm trying to support--shared state among peers.

Each peer holds its own replica, and carefully crafted delta messages are passed among peers (which can be servers, clients) to bring each other up to date using concurrent primitives (e.g. CRDTs, and sometimes simpler substrates like JSON deltas).

Mostly just thinking out loud here. It could be that the answer is "this is really not the target use case for Spin", which I can also understand. But OTOH it might also help illuminate an adjacent area of interest or future work.

[dicej]

(@dicej did that SIP end up going anywhere? Should we pick it up after the WASI P3 work is released?)

It never got past the prototype stage, unfortunately; we always had something more urgent to work on, so it was never prioritized. I did make this project as a proof of concept, but that's as far as I got.

I'd love to pick it up again, though, especially if others are interested!