A button on a monitor has a dependable background. Move it into a room and that assumption disappears. Behind it there might be a white wall, a window, another person or nothing at all. The room keeps changing while the application tries to remain understandable.
Microsoft's March introduction to Windows Holographic development makes that shift visible in its tutorial topics. Gaze, gestures, voice, spatial sound and spatial mapping sit beside the familiar work of building an application in Unity or Visual Studio. The Development Edition is approaching shipment. The interesting design problem is already clear: the physical environment participates in the interface.
Choosing a target is only half an action
Consider placing a small control on a virtual machine. A gaze direction can identify a possible target. A gesture can confirm the action. Those are separate events, and keeping them separate matters. Looking at a dangerous control should not activate it merely because looking is the only available pointing mechanism.
Feedback has to bridge the gap. The user needs to know which object is selected before confirming, whether the confirmation was recognized, and whether the application is now busy. A highlight and a short sound can do more useful work than a decorative animation if they make those states unambiguous.
The physical position of the control matters too. A panel that looks elegant in a screenshot can ask somebody to hold their head at an uncomfortable angle. A dense row of tiny targets may work with a mouse and fail with a spatial pointing method. Layout becomes a question about bodies as well as pixels.
A surface is useful, but it is not certainty
Spatial mapping suggests appealing interactions. Put an object on a table. Let a virtual ball meet a wall. Use the room to make a scene feel situated rather than suspended in front of the viewer.
But an application's interpretation of a room is still a model of the room. It should not silently turn every uncertain surface estimate into a confident physical promise. An object can indicate where it wants to be placed and let the user confirm. A scene can remain understandable while its understanding of the space improves.
This is where restraint seems productive. I would rather make one object feel convincingly attached to a place than cover every available surface with panels. Empty space gives attention somewhere to rest and makes the behavior of the remaining object easier to learn.
Sound adds another useful dimension. A quiet cue can suggest that something changed outside the current view without requiring a permanent arrow. The cue should lead to an understandable event, not make the user search the room for an unexplained notification.
A spatial interface succeeds when its relationship with the room becomes predictable. The novelty is seeing a digital object beside a real one. The lasting work is making it behave as though it belongs there.