Users are in total control of objects when they want to be, they can even take objects out of their fields.
Sometimes breaking out isn't always destrutive. But instead you break out of a field to join another. Flinging content and casting to another device is a prime example.
Sharing content is another way to break out of a field, where the physics of the direct object are shared between the two devices.
Don't make it work too hard. Make the computer work and make the objects grow and adapt their shape. We don't need to zoom in and make a user travel large distances, let's bring objects to them in a simple and direct way.
This isn't Sandra Bullock, no floating into infinite Z. There's a limit to depth per screen. ~15% of the shortest width assholes.
Respect global illumination, shadows shouldn't be going showing up and spraying everywhere. There's a reason why we don't call these things up shadows. All shadows reflect true north, even when you rotate your device.
Whether it comes from inside the touch area or outside on the hardware, touch invigorates the UI
In other cases where touch is not available, we can rely on location, shape, and form. These objects don't live -5000px in z-space, so they shouldn't scale up like they do. They instead stem from on the same z-axis and fill out into their shapes.
When something can act as a constant, a transformation can be used to ease in a new creation. It's also important to ease someone back to where they were.