But also, I agree -- foreshortening is hard until you learn 2-point perspective and can "see" perspective points that you can easily build up the foreshortened feature from.
And it behaves almost nothing like a camera. Even mechanically, the optical projection is very different than a camera makes, or what one would get from a linear perspective projection.
There is a projection, but the "image" is actually composed of multiple viewpoints over time. The adjustments an artist must make for small changes in the model's pose is nothing compared to the varying nature of the mental image you actually "see" inside your brain. It is far from static, and is in fact very dynamic.
Which is why that drawing from life is considered much more important than drawing from a photograph when learning to see.
Note that I'm not knocking the study of structure or synthetic processes like linear perspective. They are important tools and aids in perception and image synthesis. But a good drawing, or painting, or photograph for that matter, is not a mere 2D projection of light and dark. It is an encoded record of a multitude of perceptions, or perception artifacts that your brain then interprets, inside it's magic "seeing" box, which then produces a simulacrum of a perceived event or process.
Else making a good drawing or photograph would be an entirely mechanical process that anyone could make by following a few simple rules. Instead they are like moments of insight into a scene or event.
And all of this is relevant if you are creating software that is going to communicate visually. When you are building your data structures and models inside your head, it is a very different from that which the experience the user has. Your job is to make the important aspects of that model manifest visually, and usefully manipulatable.