Summaries, continued
Defining summaries on a dimension basis has a side effect of possible conflicts between the values that the cells corresponding to two or more aggregated members must be assigned.
Summation and a few other functions possess the commutativity property, so their result is invariant, but most functions would cause a conflict here. A possible solution is to give the victory to a particular dimension’s formula based on the dimension ordering. Then we would need to specify also whether this decision changes with dicing, inflations or deflations, etc.
A better way would be to assign a formula to a cell. Then we would not have to worry. However the number of such cells may be very large. An optimal solution is yet to be found.