The current format of skeletal animation in Collada seems to be not very well defined.
As it seems to me, fields BIND_POS_ARRAY, BIND_NORMAL_ARRAY, INV_BIND_MATRIX_ARRAY are excessive.
It is enough to have
a) normal and position arrays in mesh
b). “bind pose” for skeleton ( the sequence of nodes with transformations).
c). controllers for animation.
Why the current decision has been justified? It operates with too many entities.
We understand your concerns and feel the same.
For similar reasons that you state and a few others we have decided to change skinning for the COLLADA 1.4 release.
It has been a sensitive subject that has undergone a lot of debate and discussion but I believe we have come to agree on a definition that all of us find acceptable.
COLLADA 1.4 will be released in the near future. I am confident that you will find the new skinning definition less complicated and more suitable to everbody’s needs.