Wado

WEP: Target WASI P3 Only

Status

Accepted

Context

Wado is a new programming language targeting Wasm/WASI. WASI has three major versions:

The question is whether Wado should support multiple WASI versions or focus on a single target.

WASI P2 vs P3 Differences

The async model differs fundamentally:

// P2: Returns a pollable stream handle, caller manages polling
interface stdout {
    get-stdout: func() -> output-stream;
}

// P3: Takes a stream, runtime handles async transparently
interface stdout {
    write-via-stream: async func(data: stream<u8>) -> result<_, error-code>;
}

P2 requires explicit subscribe() calls and pollable management. P3's async func maps directly to Wasm Stack Switching, enabling colorless async.

Decision

Wado targets WASI P3 only. It will not support WASI P1 or P2.

Rationale

  1. Design alignment: Wado's core abstractions map directly to P3 primitives:

    • Stream<T> ↔ P3's stream<T>
    • Future<T> ↔ P3's future<T>
    • Colorless async ↔ P3's async func + stack switching

  2. Zero abstraction philosophy: Supporting multiple WASI versions would require abstraction layers, contradicting Wado's "zero abstraction to Wasm" design principle.

  3. No legacy burden: Wado is a new language with no existing users on P1/P2. There's no migration cost.

  4. Implementation simplicity: One target means cleaner compiler implementation, standard library, and documentation.

  5. Runtime support: wasmtime v40+ supports P3 with -W component-model-async=y. Other runtimes are following.

  6. Future-oriented: P3 is where WASI is heading. Designing for the past would require workarounds that compromise the language design.

Consequences

Positive

Negative

Neutral

References