One of the most underestimated features of IFC – and I believe there are many of them – is how geometries are treated.
Many BIM programs can only handle objects that have at least one (sometimes only one) geometry. In Revit, for example, you can have three detail levels for most categories and analytical geometry for some categories.
How is this different in IFC? As you can manage non geometrical objects like processes and cost items you actually don’t need any geometry for your elements. Not even for walls or spaces. While that doesn’t make sense at first (at least for designers) it gives you great flexibility when developing your virtual building.
On the other side you can assign (almost) as many geometries as you like to your elements (those derived from IfcProduct to be more precise). So a wall can have a 3D-body representation (the shape you think of when you hear “geometry”). But it can also have a 2D footprint geometry, a 2D axis, an analytical 2D plane, a center of gravity geometry (really just a point in space), an associated point cloud and many more.
Yes, there need to be applications to keep those in sync. But sometimes that is not necessary. Why use the precise shape taken from a laser scan to make a slightly wobbly “box”? Abstraction is key here.
Each shape then has to be assigned to a geometric representation context. That would allow you to switch between representations for all elements at once. Quite cool feature, but mainly overlooked.