The last commit sketches a possible structure out. I've created a new 'WebExecutor' and I've replaced the innards of Task with a PhantomData on wasm, and removed everything but the ability to forget it.

I'm not sure how I feel about this. I don't personally use the task handles, so I'm not sure if this is likely to be problematic for some users.

Thank you for starting work on this!

Quick question: how are we going to implement timers and task spawning? What crate does provide that? (sorry, not super familiar with web)

We can implement task cancelation by using something like https://docs.rs/futures/0.3.5/futures/future/fn.abortable.html

Hiya,

So timers are pretty easy, they can just be the equivalent of window.setTimeout, task spawning is... more complicated...

1. I discovered the wasm environment is not as single threaded as previously thought (though not yet as multithreaded as one would hope)
2. I had some difficulties attempting to just delegate the work to the browser APIs. The error messages from wasm-bindgen are often less than spectacular so i'm not entirely sure why.

I had another play a few hours ago and my gut feeling is that we should actually do something similar to the rest of smol, except probably in the first instance it should be limited to the thread local executor because we can't really get the full benefit of multithreading. I actually tried to just reuse your thread local executor, but it depends on io_event, which looks not so easily disentangled (for a start, it wants Async...).

So the current avenue I'm looking at involves async-task and essentially replicating the local executor, without the dependence on the IO stuff.

I have no idea what I'm doing :)

Oh, and it's web-sys that has all the relevant browser apis, but the moment you have to actually pass data across the boundary it gets a bit tedious.

Oh, one more thing I should mention, I have been taking inspiration (although not too much of it) from wasm-bindgen-futures.

Thank you for elaborating! I am currently working on untangling the reactor from the executor so this will soon become much easier.

Maybe we can use WebWorker for multithreading tasks?

Hey @jjl whats the status on this?

@kwek20 I haven't done anything on it in a while, nor am I likely to in the rest of this year.

i expect the status of the wasm backend makes some of this a bit easier now than it did when i started, if you were interested in picking up where i left off. Although given smol has changed so much since then, not sure how much of it you'd want to salvage.

Closing this PR as it is far out of date with the current master and would likely require a significant rework. Thanks anyways @jjl!

See #87 for WASM progress on specific subcrates