[gmurphy]

This is great! One very naive (I'm trying to learn) question:

You have a fixed timestep for good reason, which works great if you're displaying by doing interpolation (your display frames lag the physics and are 'catching up' to them), but in some cases where you want low latency in response to player input you might want extrapolation (your display frames are ahead of the physics engine).

I believe people typically do extrapolation by projecting forward any motion in a linear fashion, but it seems this would result in occasional weirdness like geometry intersection. One solution might be to run the physics engine at a fixed timestep for the 'real' changes, but let the graphics engine use the physics engine at a non-fixed timestep to do the extrapolation, where the results of that simulation are not taken into account for the fixed timestep frames.

Is that possible in Dyna-Kinematics, and/or am I talking nonsense? :)