Look into improving performance. There may be some gains to be had.

[brianpeiris]

Obviously performance depends on the complexity of the scene that the user decides to build. This can be mitigated by #24 but otherwise it looks like the performance bottlenecks are either inside Three.js or in the VR matrix manipulation code that we need. Hopefully the WebVR parts get optimized.

[capnmidnight]

Some of the stuff in Three.js is really good, some of it is rather naive. The raycaster is one of those naive buggers, though it provides a lot of useful data. I do a dot product test to see if my rays are anywhere near interesting objects before using the raycaster to get the detailed information. There might be similar "early, easy tests to skip hard work" patterns you can exploit.