SQL-AG approach is similar to SQL3 but it has been extended to support other applications such as online aggregation, data cube and data mining applications. Like SQL3 the aggregate will be defined in terms of iterative routines. Hence the user will define the three iterative routines for the computation of the aggregate. This requires SQL-AG to accept single (similar to initialize), multi (similar to iterate) and produce (similar to terminate) from the user. It is also required that the intermediate value of an aggregate be stored internally. This intermediate value might be a tuple of values, rather than a single value. For example in case of the aggregate count only the count of records need to be stored internally whereas, in case of average, count of records along with current sum need to be stored. So it is required that SQL-AG should allow user to specify this tuple.