class TGLCamera

 Base camera destructor.

 Set viewport extents from passed 'viewport' rect.

 Update internally cached frustum values

 Return the the current camera frustum. If asBox == kFALSE return
 a true frustum (truncated square based pyramid). If asBox == kTRUE
 return a true box, using the far clipping plane intersection projected
 back to the near plane.

 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Note: TGLBoundingBox is not really valid when filled with truncated pyramid
 - this is used as a visual debug aid only so ok.

 Return the camera eye point (vertex) in world space
 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Extract the camera eye direction (vector), running from EyePoint()
 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Find the center of the camera frustum from intersection of planes
 This method will work even with parallel left/right & top/bottom and
 infinite eye point of ortho cameras
 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Calcaulte overlap (kInside, kOutside, kPartial) of box with camera
 frustum
 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Calculate overlap (kInside, kOutside, kPartial) of box projection onto viewport
 (as rect) against the viewport rect.
 Camera must have valid frustum cache - call Apply() after last modifcation, before using.

 Calculate viewport rectangle which just contains projection of single 'face'
 of world frame bounding box 'box' onto the viewport. Note use other version
 of ViewportRect() if you want whole 'box' contained

 Calculate viewport rectangle which just contains projection of
 world frame bounding box 'box' onto the viewport. If face is
 null the rect contains the whole bounding box (8 vertices/6
 faces). If face is non-null it indicates a box face, and the
 rect contains the single face (4 vertices). Note use other
 version of ViewportRect() if you wish to just pass a static
 EFace enum member (e.g. kFaceLowX)

 Note:
    i)   Rectangle is NOT clipped by viewport limits - so can result
         in rect with corners outside viewport - negative etc
    ii)  TGLRect provides int (pixel based) values - not subpxiel accurate
    iii) Camera must have valid frustum cache - call Apply() after last
         modifcation, before calling

 Convert a 3D world vertex to '3D' viewport (screen) one. The X()/Y()
 components of the viewport vertex are the horizontal/vertical pixel
 positions. The Z() component is the viewport depth value - for a
 default depth range this is 0.0 (at near clip plane) to 1.0 (at far
 clip plane). See OpenGL gluProject & glDepth documentation

 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Convert a 3D vector worldDelta (shift) about vertex worldRef to a viewport
(screen) '3D' vector. The X()/Y() components of the vector are the horizontal
 vertical pixel deltas. The Z() component is the viewport depth delta - for a
 default depth range between 0.0 (at near clip plane) to 1.0 (at far clip plane)
 See OpenGL gluProject & glDepth documentation

 Camera must have valid frustum cache - call Apply()

 Convert a '3D' viewport vertex to 3D world one. The X()/Y() components
 of viewportVertex are the horizontal/vertical pixel position.
 The Z() component is the viewport depth value - for a default depth range this
 is 0.0 (at near clip plane) to 1.0 (at far clip plane). Without Z() the viewport
 position corresponds to a line in 3D world space - see
    TGLLine3 TGLCamera::ViewportToWorld(Double_t viewportX, Double_t viewportY) const

 See also OpenGL gluUnProject & glDepth documentation

 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Convert a 2D viewport position to 3D world line - the projection of the
 viewport point into 3D space. Line runs from near to far camera clip planes
 (the minimum and maximum visible depth). See also
    TGLVertex3 TGLCamera::ViewportToWorld(const TGLVertex3 & viewportVertex) const
 for 3D viewport -> 3D world vertex conversions.
 See also OpenGL gluUnProject & glDepth documentation

 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Convert a 2D viewport position to 3D world line - the projection of the
 viewport point into 3D space. Line runs from near to far camera clip planes
 (the minimum and maximum visible depth). See also
    TGLVertex3 TGLCamera::ViewportToWorld(const TGLVertex3 & viewportVertex) const
 for 3D viewport -> 3D world vertex conversions.
 See also OpenGL gluUnProject & glDepth documentation

 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Find the intersection of projection of supplied viewport point (a 3D world
 line - see ViewportToWorld) with supplied world plane. Returns std::pair
 of Bool_t and TGLVertex3. If line intersects std::pair.first (Bool_t) is
 kTRUE, and std::pair.second (TGLVertex) contains the intersection vertex.
 If line does not intersect (line and plane parallel) std::pair.first
 (Bool_t) if kFALSE, and std::pair.second (TGLVertex) is invalid.

 NOTE: The projection lines is extended for the plane intersection test
 hence the intersection vertex can lie outside the near/far clip regions
 (not visible)

 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Find the intersection of projection of supplied viewport TPoint (a 3D world
 line - see ViewportToWorld) with supplied world plane. Returns std::pair
 of bool and vertex. If line intersects

 Camera must have valid frustum cache - call Apply() after last modifcation, before using

 Apply a 2D viewport delta (shift) to the projection of worldRef onto viewport,
 returning the resultant world vector which equates to it. Useful for making
 3D world objects track mouse moves.

 Camera must have valid frustum cache - call Apply()

 Calculate if the an object defined by world frame bounding box
 is 'of interest' to the camera. This is defined as box:

 i) intersecting completely or partially (kInside/kPartial) with
 cameras interest box (fInterestBox)
 ii) having significant length OR volume ratio compared to this
 interest box

 If a box is 'of interest' returns kTRUE, kFALSE otherwise.
 See TGLCamera::UpdateInterest() for more details of camera interest box.

 Note: Length/volume ratios NOT dependent on the projected size of box
 at current camera configuration as we do not want continual changes.
 This is used when (re) populating the scene with objects from external
 client.

 TODO: Might be more logical to move this test out to client - and
 have accessor for fInterestBox instead?

 Update the internal interest box (fInterestBox) of the camera.
 The interest box is an orientated bounding box, calculated as
 an expanded container round the frustum. It is used to test if
 if object bounding boxes are of interest (should be accepted
 into viewer scene) for a camera - see TGLCamera::OfInterest()

 The interest box is updated if the frustum is no longer contained
 in the existing one, or a new one calculated on the current frustum
 differs significantly in volume (camera has been zoomed/dollyed
 sizable amount).

 If the interest box is updated we return kTRUE - kFALSE otherwise.

 Clear out the existing interest box

 Adjust a passed REFERENCE value 'val', based on screenShift delta.
 Two modifier flags ('mod1' / 'mod2' ) for sensitivity:

 mod1 = kFALSE, mod2 = kFALSE : normal sensitivity (screenShift/screenShiftRange)
 mod1 = kTRUE,  mod2 = kFALSE : 0.1x sensitivity
 mod1 = kTRUE,  mod2 = kTRUE  : 0.01x sensitivity
 mod1 = kFALSE, mod2 = kTRUE  : 10.0x sensitivity

 'val' is modified and clamped to 'min' / 'max' range.
 Return bool kTRUE if val actually changed.

 Used as common interaction function for adjusting zoom/dolly etc

 Draw out some debugging aids for the camera:

 i) The frustum used to create the current interest box (RED)
 ii) The same frustum as a squared off box (ORANGE)
 iii) The axis aligned version of the frustum used as interest box basis (YELLOW)
 iv) The current interest box (BLUE)

 Return one of the planes forming the camera frustum

 Camera manipulation interface (GL coord - origin bottom left)

 virtual void   Frame(const TGLBoundingBox & box) = 0; // TODO
 virtual void   Frame(const TGLRec & rect) = 0; // TODO

 Window to GL viewport conversion - invert Y