In order to minimize impact of rounding errors, we need the 
resolution we do our layout at to be significantly higher than the 
resolution at which we will produce the output.

The issue I am not entirely certain about is whether (a) this system of
units should be used throughout our code and translation to screen 
units
be made wherever system API is accessed, or, (b) the internal units 
should
only be used inside the layout and graphics classes, while the GUI 
classes
(the frame, toolbars, ect) would continue to use the system units.

The advantage of (a) is consistency, the main advantage of (b) is 
avoiding unnecessary converions within the GUI code, as well as not 
needing to interfere with the present GUI code, save at the 
boundaries between it and our layout and graphics classes. I 
personally am slightly inclined toward (b), but have not made up my 
mind and would appreciate some input.

Tomas

Reply via email to