Navigation - Too much freedom?

I'm currently writing a paper about the use of GEarth Tours in education.  I thought I'd discuss one aspect that's come up: the problems of navigation in 3D software (thats Maps as well as GEarth as you effectively move through 3D space using pan and zoom).

The Problem: In GEarth you have 5 degrees of freedom:
  Altitude, Latitude, Longitude, Camera Pitch, Camera direction

Controlling these is complex and so causes problems - users can:

• become lost in virtual space
• get disorientated 
• become confused as they fly through model walls (ie inside a building only designed to be seen from outside)
• navigate around missing the views that the designer of a GEarth project wanted them to see.

As you may imagine any combination of these leads to the same result:  users become disheartened and leave to go and look at something else.

Similar Problems in other Software:  It isn't just GEarth - map systems, information spaces with pan and zoom functionality and Virtual Worlds all suffer from similar problems.  In some software its possible to fly straight into the ground with nothing visible at all which is when you get 'desert fog', users don't know where they are with no visual clues on screen to help them.

Solutions:  There are multiple solutions I've found in the literature, one is illustrated by Google Body:  Constraining freedoms of movement from 5 to 2.  When viewing the body your camera angle is fixed and you can only fly around a cylinder of fixed distance from the body (in fact it feels more like you are moving the body rather than your camear position but actually they add up to the same thing).  

GEarth Tours:  The solution readily available in GEarth is the GEarth tour, effectively you are constraining the user to 1 degree of freedom as within a tour they only have the ability to play or rewind  - nothing else.  This means that providing the tour is designed well we mitigate or solve all the problems I listed above.  

In education removing those issues comes at a cost:  users watching a tour are far more passive than if they are navigating around so we have to be careful to insert active tasks into tour.  For example, turning the tour off for a while and having students go and search for a landscape linked to the GEarth tour they've just viewed.  How to do this is the theory that we are investigating with our Google Research Award.