Changing switch -R <num> in &gla to mean the max allowed size of abstraction. Adding switch -Q <num> to stop when the number of objects exceeds num % _during_refinement_.