There's an identical question here but it was never answered fully and the link providing an essential component of the accepted "answer" is broken. The Keisan website presents a solution here (PDF) that shows the length of an arc going clockwise from the Y axis as $${s(x) = aE({x \over a},k)}$$
Consider an ellipse with a = 20, b = 5 (k = 0.9682) and an arc starting at (0,5) and ending at (7.3625, 4.6489). E(0.3681, 0.9682) = 0.3604, according to the Keisan integral calculator, yielding an arc length of 7.208. But using Keisan's own arc length calculator (the angle is 32.27°) yields a length of 7.3736. Close, but not exact.
For an arc ending at (18.8774, 1.6516): E(0.9439, 0.9682) = 0.8196 yields a length of 16.392 versus the calculated 19.3729 (angle of 5°). Considerably off.
Keisan certainly seems to have done something right in their arc length calculator! But I can't duplicate their results, so I'm clearly missing something in the calculation of the parameters and/or the application of the formula. Can anyone provide a clue, or suggest a concrete path for how to apply arc parameters such as these to the elliptic integral to calculate the arc length?