[moron4hire]

Oh man, I feel you.

I think part of the problem is that the full understanding of a scene with interactions requires a graph structure, but most of our tools are really only built for editing tree structures.

I see this time and again with just regular, ol' 2D DOM and JavaScript. You can define your layout in HTML and then play whack a mole with bugs if elements you got via document.querySelector move or get renamed, or you can define your layouts in JS and pay a relative huge cost in rendering and complexity.

Plus there is the added wrinkle in 3D of having to be very persnickety about memory and shaders and order of operations. How do you make a declarative, semantic scene description that can figure out that this geometry over here can be batched, this over here can be instanced, and never the twain shall meet?