Regardless of a few complaints here and there about fixed frames, in my opionion Blender has performed very well for scientific kinematic modeling and visualization of meshed data. Most complaints about fixed frames are made without an understaning of numerical discretization, which is part of most high-end scientific calculations.
There are a few non-deal-breaker issues that I think should be addressed.
(1) Although Blender animation is based on fixed frames, ipo curves are a function of continuous Blender time which makes scientifc kinematic modeling and visualization possible (by insertion of ipo points with Python). However, the fixed-frame nature of Blender does involve one unfourtunate property:
One cannot access transformation matrices between frames even though ipo curves and points may be present at intermediate positions. The user must shift all ipo curves so that the intermediate point now lies on a frame. Then and only then is extracting intermediate rotation info possible (I think).
Is there a way to access intermediate rotation info that I am missing? I am looking for something like:
Blender.Set(‘curtime’, time)
rot_mat = obj.mat
How difficult would it be to add this in the source code? I have a feeling that it would not be too difficult since ipo curves are defined at all times.
(2) During shape key deformation, relative motion between points falls along a line that connects intial and final points. If the time step is too large, this linear interpolation may not be adequate for motion along a non-linear surface (e.g. a sphere). One workaround is to use smaller time steps between shape keys. However, it would be nice if motion could be non-linear. I would like to, for example, deform a spherical cap while having vertices stay at a constant distance from the center of the sphere. Any ideas?
(3) Blender needs Delaunay triangulation !!! Yes, I can use code from the web or write my own. Option 1 is associated with copyright issues beyond Blender. With respect to option 2, efficient and robust Delaunay is non-trivial and should be contained in the fast C-source. If delaunay triangulation was added to Blender, I think many more scientific modelers would be hooked!!! This would also be useful for artistic stuff as well.