Feel like if we ever want to break the browser hegemony we need a reboot of web technology to start clean with a design that's actually feasible to implement from scratch without spending billions of dollars
I'm sure someone will reply to this with a smug "oh *just* simplify the web huh?" implying it can't be done but honestly web technology as we know it today is a janky blob of different things, many of which were given basically no thought when they were initially added, all bolted onto one another
Like, do we really need all of SVG, Canvas2D, WebGL (1 *and* 2 which aren't mutually compatible!) and WebGPU? Really? Do we need that? Or could that be simplified. And that's just one example
serious answer
@eniko Annoyingly, the answer there is *sort of* yes. WebGL 1 and 2 could conceivably be deduplicated, given enough time and GPU driver updates, but WebGPU is built on Vulkan instead of OpenGL and is AFAIK simply not possible to support on older GPUs.
All of those require hardware acceleration to be usable; for those cases where that is not available, there's Canvas and SVG which have almost no overlap in functionality; one is a raster graphics API, and the other is a vector API.
Could these APIs conceivably be merged? Yes, but the total API surface wouldn't look much different from what the combined API surface is right now, so in terms of implementation complexity it wouldn't really matter.
Assuming that breaking backwards compatibility is acceptable, there are definitely a lot of things in the various browser APIs that could realistically be replaced or merged with a design iteration (think fetch vs. XMLHttpRequest), but I suspect that graphics APIs will not be one of those...
serious answer
@joepie91 @eniko There's a subset of WebGPU that works with OpenGL ES 3 (at least with the dawn engine). Once I migrate to WebGPU I will see if I can just delete all my GLES code without loss in compatibility/performance.