Jasper St. Pierre

Hmm, this is likely an issue with the web CacheStorage API, also related to #550.

I'd need more RE work done on the game to know what to do here.

I assume the actor has custom code for drawing this model. Would have to look into what it is.

That definitely sounds odd. My guess is that it's something to do with pointer lock. Can you try running `main.viewer.inputManager.usePointerLock = false` in the console and see if that improves...

Looks like an issue with Firefox's pointer lock implementation on Linux -- I think it might be ignoring pointer acceleration. There's not much I can do about that.

It's mentioned in the Controls section on the README, but I can probably find a better place for it: https://github.com/magcius/noclip.website#controls

Bizarre that Intel does not have clip planes, since it's required for D3D10+ through the SV_ClipDistance semantic.

For a single clip plane, consider using an oblique projection matrix: http://terathon.com/code/oblique.html

I suspect a large portion of flat interpolants are cases like passing `instance_id` from the vertex shader to the fragment shader, or passing some kind of draw-uniform data across. While...

> Is that always the case or are there plenty of cases where keeping things in a vec3 or vec4 has performance advantages but then I lose any semantic meaning....