Millimeter-wave (mm-wave) and sub-terahertz (sub-THz) communications are expected to be one of the biggest beneficiaries of the emerging reconfigurable intelligent surface (RIS) technology. RISs can compensate for large path loss and blockage inherent to the mm-wave and sub-THz frequency bands to yield enhanced communication performance in these bands. To achieve high beamforming gain and realize the enhanced performance in RIS-assisted wireless communication, the acquisition of accurate channel state information (CSI) is critical. In this article, we provide an overview of channel estimation for RIS-assisted mm-wave/sub-THz communication to address technical challenges, tradeoffs, channel estimation frameworks, and training signal design. We summarize the recent RIS-related sparse channel estimation approaches based on beam-space, sparse recovery, array signal processing, and data-driven techniques, highlighting several challenges for future research.