Let's say instead of x2-x1, we say, "The difference between x2 and x1", let's call this value A. We'll do the same for the y variables and call that difference B, for simplicity.
For (2,3) and (4,9), we're really dealing with those differences, so A should equal 2, and B should equal 6. So far so good. But this is pretty much the "rise and run" of a triangle. 2 to the right, then 6 up. But if you walk 2 miles east, then 6 north, while you, personally, have travelled 8 miles, the distance between your starting point and where you are is NOT the way you walked.
So in the case of the triangle, using 2 as the run and 6 as the rise, to get the straight line between the starting point and the ending point, we have to do Pythagorean's Theorem, or A2 + B2 = C2 to get that last direct line between the two points. Cancelling the root and exponents removes the process to get that exact distance squared. We then take the square root of that to get the exact distance.