I'm aware there's a Wasmtime fork for now, but I'm curious about the current thinking on how to "ship" a wasi-gfx implementation for Wasmtime longer-term.
I ask specifically because my understanding is that Wasmtime-the-project doesn't necessarily want to carry implementations for all WASI proposals in its own tree. (I don't remember where I got this idea, so it may well be flatout wrong!)
Is there some semblance of a plan for how users would eventually use a wasi-gfx implementation with Wasmtime? E.g. upstream one into Wasmtime? Use wrpc if that can be fast enough? Maintain a forever-fork?
I understand that this whole space is in flux — I'm just curious about current thinking on it.
I'm not associated with Wasmtime, so don't know what the plans there are. However, I do maintain wasi-gfx-runtime, which can be added to Wasmtime (and in future to other runtimes as well) quite easily.
We have had some discussions about merging into Wasmtime a while ago, but we didn't feel ready at the time, so we didn't push for it from our side.
I don't know what the Wasmtime maintainers think about this, would love to hear as well.
FWIW, as I expressed in previous offline conversations, I think there's a lot of openness to having wasi-gfx support upstreamed into wasmtime proper. I personally would be very excited about that, and would encourage pursuing it whenever you feel ready to do so!
Last updated: Jan 24 2025 at 00:11 UTC