<jpab> basically what we need to do is: 1) give transform nodes an ID; 2) refer to them by ID rather than pointer in the AnimationChannel; 3) turn Model into ModelInstance, which holds a pointer to the static node graph and a vector of cached transforms; 4) shift & rework Animation::Interpolate to take the static data and a time and interpolate it to give a transform that gets put into the model's transform cache
<jpab> I guess we actually only want to give IDs to nodes that have animations, 'cause there are static transforms too
<jpab> and since there's no blending, there's a one-to-one relationship between animation channels and animated transform nodes
<jpab> so actually transform nodes can know what their animation channel is
<jpab> but maybe we want to support blending in the future, I don't know
<jpab> massive advantage of having models as data instead of scripts: we can keep rewriting and improving the model rendering code without breaking all the models every time