A modeling transformation converts an object's 3D coordinates, specified in a local "object" space, to the "world" coordinates of the entire model. This is performed using a simple matrix transformation, so what could go wrong?

However, consider the situation where you are displaying a simulation of a tsunami in the Indian Ocean. You might find that the simulation freezes or disappears, while tests of smaller subsets of the data look fine. What might be happening?

OpenGL's internal transformations work in a space scaled between -1 and 1. If you pass it data measured in centimeters for a tsunami spanning the Indian Ocean, the data dimensions could cause an overflow. The solution in this case might be to allow earlier stages of the visualization pipeline to work in a normalized coordinate system with more reasonable units.