This article tries to close the long-run debate of the evolution of the digital divide, taking the Internet and cellular phone as two examples. We first compared the sensitivity of four algorithms of the digital divide, including the standard deviation of the adoption rate in each subgroup or country, entropy of the adoption rate, Gini-style Coefficient and Zhu's algorithm. It is found that the variance-based measurements were more sensitive than the proportion-based algorithms. We then illustrated the worldwide and within-society evolution of the digital divide with these four algorithms. With the growth of the adoption rate, we found an inverse-U shaped trend of the dynamic of the digital divide of both the Internet and cellular phone, which may indicated that with the growth of the information technologies adoption, the digital divide would widen at first, and then naturally narrowed.