OpenCV's cornerScore is different from FAST paper

Question About Fast9 Score
v is the brightness of center pixel, v_i is the i-th pixel brightness on circle. OpenCV use this code to calculate Fast9 Score.

\begin{align*} & D = \{v - v_i\}_{i\in[0,24]}\\ &\text{for i=0 : 15} \\ &\{ \\ & \quad\quad d_{min} = \min(D_{i+1}, D_{i+2},...,D_{i+8})\\ & \quad\quad d_{max} = \max(D_{i+1}, D_{i+2},...,D_{i+8})\\ & \quad\quad q_{min}= \max(q_{min}, \min(d_{min}, D_i) )\\ & \quad\quad q_{min} = \max(q_{min}, \min(d_{min}, D_{i+9}) )\\ & \quad\quad q_{max}= \min(q_{max}, \max(d_{max}, D_i)) \\ & \quad\quad q_{max}= \min(q_{max}, \max(d_{max}, D_{i+9})) \\ &\} \\ &\text{score} = \max(q_{min}, -q_{max}) - 1 \end{align*}

However, in the paper: Machine learning for high-speed corner detection, the Score is:

V = \max(\sum_{x\in S_{bright}}|I_{p->x}-I_p|-t,\sum_{x\in S_{dark}}|I_p-I_{p->x}|-t )

1、Why not use the method of paper?
2、Which paper is the source of OpenCV’s method? OpenCV’s method is better?

No one reply me. Finally i find the answer in c++ - Algorithm behind score calculation in FAST corner detector - Stack Overflow