As already known commonly, high-spectral-resolution lidar technique (HSRL) employs a narrowband spectroscopic filter to separate the elastic backscattered aerosol signal from the thermal Doppler broadened molecular backscattered contribution. This paper presents a new and comprehensive view of HSRL technique from the perspective of spectral discrimination, without concretizing the analysis into a specific spectral discrimination filter. Based on a general HSRL layout with three-channel configuration, a theoretical model of retrieval error evaluation is introduced. In this model, we only take the error sources related to the spectral discrimination parameters into account, and ignore other error sources not associated with these focused parameters. This theoretical model is subsequently verified by Monte Carlo (MC) simulations. Both the model and MC simulations demonstrate that a large molecular transmittance and a large spectral discrimination ratio (SDR, i.e., ratio of the molecular transmittance to the aerosol transmittance) are beneficial to reduce the retrieval error. Moreover, we find that the signal-to-noise ratio (SNR) and SDR of the lidar system are often tradeoffs, and we suggest considering a suitable SDR for higher molecular transmittance (thus higher SNR) instead of using unnecessarily high SDR when designing the spectral discrimination filter. This view interprets the function of the narrowband spectroscopic filter in HSRL system essentially, and will provide some general guidelines for the reasonable design of the spectral discrimination filter for HSRL community.