(1) We accept the modelling (geometry) graphics categorization and consider our area to fall within modelling (geometry).
(2) We restrict transformations to 2-D and 3-D affine and clipping (windowing.) Note : Virtual viewports can be defined with affine transformations. A window or clipping region is defined as a set of points, e.g. the set of points inside (and on the boundary of) a square.
(3) We assume that pictures are structured hierarchically. We assume the following details of the structure:
i Gi Wi G W j Gj Wj
Then two possibilities exist
(a) Clip before transform:
Gj = Gi O Gj where O stands for composition Wj = Gi-1 (Wj) I Wi where I stands for intersection
(b) Transform before clip:
Gj = Gi O Gj where O stands for composition Wj = Gi Wj I Wi where I stands for intersection
Transformations premultiply pictures here
Case (a) has the advantage of not transforming points that will later be clipped; the disadvantage is that the inverse Gi-1m must be computed.
Case (b) does not involve inverse but must transform all points before clipping.
Conclusion (tentative) : both orders should be permissible; a particular system could select either order or both.
(4) The leaf nodes of the tree contain calls to virtual (world model) picture primitives. If p is a primitive, then the node:
i:p Gi Wi
means either:
Draw virtual (Gi(p) within Wi) (case (b) above)
or
Draw virtual (Gi(p within Wi))(case (a) above)
Note : Therefore transformations on a picture = transformations on primitives)
Draw virtual constructs the world model picture and thus defines part of the interface to the Graphics component.
See above for possibilities.
Editors note: The report of this sub-group can best be understood with reference to the paper Concatenating Transformations in Hierarchic Graphics, W.R. Mallgren and A.C. Shaw, (Technical Report No. 76-02-2), January 1976, Department of Computer Science, Univ. of Washington, Seattle.