Memory-based decision tree algorithms, such as C4.5 and its derivatives, do not support scalability well as they depend on the available memory. If the dataset being classified is very large, the efficiency also suffers as a large size of cases has to be traversed many times. Fortunately, object-relational DBMSes offer features (such as indexing,query optimizer, improved SQL and stored procedures) to speed up queries and business logics that could be used in implementing the algorithms. This paper presents features that would be useful in improving efficiency and scalability of the algorithms. On the last section, we also outline our purpose technique that utilizes the features and the result of early experiments.