@ Jefferytitan My goal is to create a tree with physical objects at the bottom most level or leaf nodes. The idea behind voxels is that I will only draw the faces where one block is filled and it's neighboring block is empty. In this way I'm not drawing lots of extra faces and killing performance. So the upper part of the tree will be conceptual and the bottom will actually have the mesh and colliders and all. I'm thinking that I might just keep all of the tree conceptual and instantiate the prefabs separately.
@Slayemin I have read your article, and while it was useful for object collisions what I am working towards is using the octree structure to draw the blocks in the world.
I am actually working on implementing a different kind of voxel that is a triangular prism with equilateral triangles for top and bottom and squares for sides( I have a simple Html5 demo of what I'm trying to do at http://apoidgames.web44.net/wedge.html). So I won't be able to use the cube prefab or cubic shapes in general. To accomplish this I have adapted the cubic coordinate system for hexagons from http://www.redblobgames.com/grids/hexagons and made it 3d. Basically I am replacing the X and Z axes and replacing them with Q, R, and S axes, which I put together another simple 2d demo of at http://apoidgames.web44.net/trigridX.html to show what I mean. One of the coordinates is redundant, which means I could represent all of the points with only 2 of these axes, but because of this redundancy I have the attribute of for every QRS value the sum of all the values will always equal 0 ( which has some very useful properties). In the end I will have QRSY coordinates. The octree data structure will allow me to create space partitioning on which I would be able to create the world of what I call triblocks. I could accomplish the same thing using a 4 dimensional array, three dimensions for the XYZ coordinates and 1 more to split each block in half, but I think that an octree would be more efficient way to store the world.
My biggest problem is that my university training has given me a ton of theory and not a lot of practice, so I know enough to know that this would be a good way to do this, and not enough to be able to implement the theory. So that is why I was looking for some help from someone who has implemented the octree for the purpose of voxel rendering to help me out. I may end up implementing this in XNA instead since I have tutors here at school who know that system better. I just don't like the idea of implementing my own collisions and also being so limited as far as what platform this would run on.