Can you say more about location?

[nedbat]

The spec mentions bounding boxes, and shows examples of location elements:

<location value="10"/><location value="20"/><location value="30"/><location value="40"/>

A few questions:

• Why does an AI need to know the bounding box?
• Why is a format focused on semantics even concerned with where the information might land in some rendering? Don't renderings differ?
• Why use this odd four-element format? Why not <location x_min="10" y_min="20" x_max="30" y_max="40" />

The spec even says (about other formats):

Whereas these formats are primarily concerned with the geometric locations of the various spans of text,

But HTML does not explicitly specify the geometric location of text, while doclang does.

I guess there are many things about location I don't understand.