Non-orthogonal multiple access (NOMA) allows multiple users to share a time-frequency resource block by using different power levels. An important challenge associated with NOMA is the selection of users that share a resource block. This is referred to as clustering, which generally exploits the channel disparity (i.e. distinctness) among the users. We discuss clustering and the related resource allocation challenges (e.g. power allocation) associated with NOMA and highlight open problems that require further investigation. We review the related literature on exploiting channel disparity for clustering and resource allocation. There have been several misconceptions regarding NOMA clustering including: 1) clustering users with low channel disparity is detrimental, 2) similar power allocation is disastrous for NOMA. We clarify such misunderstandings with numerical examples.