In recent years, due to the increasing popularization of data broadcasting, the volume and variety of data being broadcast are rapidly increasing. In this environment, as it is di cult for users to search for information from a large amount of broadcast data, there is an increasing demand for ltering techniques that automatically extract only the necessary data. Consequently, a number of ltering methods have been proposed. However, mathematical representation of these methods does not exist. T h u s , i t i s n o t possible to qualitatively evaluate various ltering methods, optimize processing methods in ltering, or design a declarative language for ltering processes. In this paper, we de ne ltering as a function, and express the properties of ltering methods by the constraints satis ed by this function. By showing the inclusion relation of constraints representing the properties of ltering, we clarify the relationship between the properties of ltering. Using the framework proposed in this paper, we are able to categorize the actual ltering systems. By applying the appropriate processing method for each category according to its properties, more e cient ltering systems can be achieved.Permission to copy without fee all or part of this material is granted p r ovided that the copies are not made or distributed for direct commercial advantage, the VLDB copyright notice and the title of the publication and its date appear, and notice i s given that copying is by permission of the Very Large Data Base Endowment. To c opy otherwise, or to republish, requires a fee and/or special permission from the Endowment.