What the user intends to do with the structure can influence what is actually meant by a fair division of computation. Since we cannot predict this we assume that the bulk of the operations will be traversals around the structure, with occasional adding and modification of data values. In the situation where we merely want to map a function over our data structure, dividing the data out equally between the processors is more important than the locality of the parts of the structure. It becomes more important when the function (or amount of work to be done) increases, and dominates the actual overhead of traversing the tree.