A few PLSM questions

I have a couple of questions concerning the PLSM

1.)
Is their an optimal method of distributing the scene graph such that heirarchical occlusion culling achieves optimal results? (example, If one is in a room shoud the room be an object with the camera as a sub node so that the room is rendered first, followed by other rooms around it, followed by the terrain (which will then only be rendered if it can be seen through any windws)

2.)
On texture mapping, I am under the impression that if I want to achieve the greatest quality texture (bump mapped landscape, noise, possible texturing similar to terragen's material system, etc) I will need to rewrite the PLSM shader (ie I cannot simply specify ogre material and shaders and plug them in to each texture slot). Is this correct?

3.)
Can pages for the terrain system be manually specified, and their data expounded on, as opposed to using the mapsplitter? Clearly there is far more information to a page than simply the terrain. Vegetation, structures and other such information should be loaded also (at least the program should be made aware of their existance so that the proper preparations may be made)

4.)

Finally, thanks!

1.) Optimization is about ratio on what is to be queried (GPU render cost / GPU query Cost). Rules are defining big occluders, group Occluders/ occludee into static geometry, etc... to make the ratio more interesting.

2.) Best is to have to rewrite a new TextureMode, using existing example (texturemode and associated shader+material) as basis. But "Image" texture mode supports using material like "paginglandscape.X.Z" X and Z being page coordinates.

3.) using "page listener" mechanism, you get a preload event when a page is loaded, and that's where you can get data and/or load your own.