As a most successful technique for building recommender systems, collaborative filtering has been the foundation of many commercial systems. Major idea of this technique is searching many other users having similar preferences with the current user, referring to their rating information of items, and recommending items that might be preferred by the current user. This paper examines the characteristics of the current collaborative filtering systems with research or commercialization purposes and investigates various fundamental issues related to collaborative filtering techniques, which would assist in the understanding and future research of collaborative filtering.