animation stack data

I am exporting data that looks very similar to collada’s matrix-stack and associated stack data. I am concerned about the efficiency of this data and it’s size. Can anybody describe why collada went with this matrix-stack format and pros and cons of using it as is?

I’ve always used local quaternions/translations/scales for each keyframe per joint as opposed to this matrix-stack approach. I’m not making a judgement here that my way is better, I just want to open a discussion on this topic.


A general matrix stack is more flexible than a fixed set of transforms. In general, the Collada designers seem to prefer flexibility to implementation efficiency. I think that’s for the best.