A comment on the double-chained tree
In 1963, Sussenguth [8] suggested that a file should be organized as a double-chained tree for searching and updating. Patt [5] then obtained the optimum double-chained tree under the assumption that no key may prefix another and that all terminal nodes (items of information) have equal probabilities of being searched. Stanfel [6, 7] explored the relation between the double-chained tree and variable length code [3] and solved a special integer programming problem which corresponds to the case of equal probabilities of terminal nodes and a finite set of available symbols for keys with different costs.