Explore separation of interaction system into logical/physical

[SnowyCoder]

Now the interaction system handles all kinds of interaction, but that might not be the best thing as it requires a lot of hacks and workarounds.
As an example: pins are logically points with no dimensions, but they should be selectable. To do this the clicks used to query the selected entities are queried as circles. But if you look at things like background images they should not be queried with circles (if you click near an image, but not in the image, it should NOT be selected).
So what, you may naively ask, just make the pin a circle instead of a point.
This is even more destructive as systems like the visibility system query the interaction system to know if a pin is visible by a visibility polygon.

The thing I'm exploring as the "perfect solution" is to divide the spaces in two:

• A logical space, which represents each item as it is in code (pin -> point)
• A physical/interaction space, which represents each item as the interaction shape (pin -> circle)

But will this really help? What should each entity be in the physical space (wall -> two circles connected by an OBB?).