As far as I can tell, that plugin is using a uniform voxel system, as opposed to the sparse octree approach I’ve used here. Based on that :
DoN (Voxel) Advantages:
- Simple data structure
- Easy to update partial sections of the navmesh
- Good for small, dynamic levels
SVO Advantages
- Much more space efficient
- Vastly improved A* performance across long distances
- Good for huge, static worlds
Not that you couldn’t mitigate some of the disadvantages of each…but that’s roughly accurate. The DoN approach will die if you want to path across long distances, the A* iterations will kill you.