Going back to my photography school days almost 30 years ago. A "normal" lens is defined by the film size or sensor size. That is the diagonal of the sensor size in mm is equal to the "normal" lens size in mm.
Using the Pythagorean Theorem which is a square + b square = c square. Therefore the APS-C sensor on the 7D for example which is 22.3 mm x 14.9 would result in a normal lens of 26.82 mm.
A full frame sensor would have a "normal" lens of 43.27 mm.
This is a rather dogmatic view of what "normal" means. "Normal" is supposed to mirror human's field of view, which is close to fisheye with terrible resolution off center, and long telephoto with good resolution. When comparing different aspect ratios, comparing the FL to horizontal or vertical size becomes inconvenient; then the diagonal is more convenient as some kind of average of the two. One can also take the square root of the area instead.
Once we decide that the we prefer the diagonal as a linear measure of the size of the film/sensor, then FL=diagonal is as "normal" as FL=1.12345*diagonal. Choosing it to be exactly equal to the diagonal sounds a bit mystical, so it must be right.
If you really want to know what "normal" means, this can only be determined with experiments with people. It is tricky because the AOV depends on the distance to the print, and the latter would depend on how large the print is, in a non-linear way.
Yes, I know what wikipedia says.