FWIW the vfx industry is starting to settle on a single format (again) - this might make more sense to implement than something more xml/web-ish... https://graphics.pixar.com/usd/docs/index.html

USD couldn't be more different than X3D. It's an industrial-strength standard for sharing massive movie-scale assets between commercial editing suites and production tools. X3D is lighter weight markup for defining simple 3D scenes that can be viewed in a web browser with a little interactivity.

How does it work? Does it just render a scene from server-side?

I've just seen X3DOM which as I understand, it has something called "explicit DOM" means that it loads different DOMs explicitly which is useful I think, especially when the scene is complex

I'd much rather prefer A-Frame become this layer for HTML, if only for the fact that the component registry is already community driven and therefore populated.

this could be quite powerful if combined with react. excited to watch it develop!

See this relevant project[0] which enables you to use JSX/React to program the popular 3d library three.js

[0] https://github.com/Izzimach/react-three/blob/master/README.m...

X3D itself predates both React and A-frame, there is even X3DOM to use it with WebGL: https://www.x3dom.org/