That manual is for legacy (now unsupported) versions:
4.X and older. Current version is
6.0.1.
Since version 5.0, simulation is always performed in the solver's local space. This results in better performance and makes it easier to reason about what space you're working in. It also avoids floating point precision problems when working away from the world's origin, which is a must for large game worlds. Quoting the manual:
http://obi.virtualmethodstudio.com/tutor...olver.html
Quote:Solvers always perform the simulation in local space. This means that translating, rotating or scaling the solver will rigidly transform the simulation as a whole.
To emulate world space, just leave the solver at the world's origin. This way, local and world space become the same. Note you can also inject world-space inertial forces into the solver, using the solver's
world linear/angular inertia scale parameters.