This is all about projective geometry. For common lenses (rectilinear lenses, approximated by the Pin-hole-camera model), parallel lines in 3D converge into single points in a 2D projection. (By definition those lines that stay parallel in 2D, too are considered to intersect at infinity.)
The book to read would be Multiple View Geometry by Hartley and Zisserman.
The technique you describe revolves around having a very long focal length. If you are happy with a bit of Mathematics, the essential part of the Camera Matrix is that y1 = f / x3 * x1 in the image below.
What you want to do in this image is make the green line horizontal. By increasing f and x3 the angle between the horizontal axis X3 and the green line (the ray the light takes) becomes smaller and smaller.