Make go-libp2p compile to WASM

[v1rtl]

Makes go-libp2p compile by omitting pion/webrtc which explicitly does not support WASM for certain APIs.

Fixes #3277

Additionally

• all builtin transports are disabled for WASM in default settings, as they are not supported by browsers
• Some utilities that are still used elsewhere are moved into *_shared.go files

I locally tested and verified that both linux and wasm targets build properly

This is a more basic approach to #3280, whose goal is to bring WASM support to WebRTC, but it's blocked by upstream dependencies

[MarcoPolo]

Thank you! I think this would be great to merge. For reference, did you see #2191?

I'll try to review this this week.

[MarcoPolo]

As a rough guideline, I think this would be a good way to merge Wasm support:

1. Make changes to be able to compile go-libp2 without any transport.
2. Add transports support in one by one.
   2a. Extend libp2p's transport interop tests to test the transport. (I can help with the specifics here)

[v1rtl]

Thank you! I think this would be great to merge. For reference, did you see #2191?

I'll try to review this this week.

I didn't know about this PR, neat! Since then a lot of things have progressed, browsers support WebTransport much better now and Go has better WASM interop

[v1rtl]

As a rough guideline, I think this would be a good way to merge Wasm support:

1. Make changes to be able to compile go-libp2 without any transport.

Do you mean to not include any transports by default for wasm target?

[MarcoPolo]

As a rough guideline, I think this would be a good way to merge Wasm support:

1. Make changes to be able to compile go-libp2 without any transport.

Do you mean to not include any transports by default for wasm target?

Yep. Just to get everything else to compile.

[v1rtl]

As a rough guideline, I think this would be a good way to merge Wasm support:

1. Make changes to be able to compile go-libp2 without any transport.

Do you mean to not include any transports by default for wasm target?

Yep. Just to get everything else to compile.

well this PR already makes it compile, despite those transports not working in WASM yet

do you still prefer if I remove all transports from a default config anyway?

[v1rtl]

Explicitly disabling wasm compilation for some unsupported APIs managed to remove 6 megabytes from the binary, mainly due to exclusion of quic-go

[pasha1980]

@MarcoPolo Any update on that? I assume it's on hold?

[MarcoPolo]

I think the core go-libp2p can compile to wasm. It's more about building a useful transport within wasm.

Some work here: #3386 that facilitates building libp2p hosts without needing packages that can't compile to wasm.

[v1rtl]

I think the core go-libp2p can compile to wasm. It's more about building a useful transport within wasm.

Some work here: #3386 that facilitates building libp2p hosts without needing packages that can't compile to wasm.

it can't last time i checked. fails because of transport libraries that get imported as a side effect. I still have to patch stuff in go-lib2p to make it compile without trying to pull non-wasm-compatible stuff

[pancsta]

Id be very interested in this - currently working with WebSockets, WebWorkers, MessageChannels, and TCP relays. WS and worker ports are represented as plain net.Conn. You can take the same approach as I did and inject an existing connection (tcpreuse transport?). I also have a libp2p-pubsub integration and would love to bring it to the browser.

Useful links

• working demo
• WS-netconn bits
• WS-to-TCP relay