Review of "Hierarchical data-base management:: a survey" by D.C. Tsichritzis and F.H. Lochovsky.

Thls survey provides a useful introduction to some of the facilities of IMS (which I know something about) ; it appears to be an equally useful introduction to System 2000 (which I know nothing about). I do wish, however, that (I) the paper had been more successful in its attempt to provide a general motivation for hierarchical data structures, and (2) the paper had @rovided a more complete overview of all aspects of the systems. In particular, in spite of the central role accorded data description in Sibley's introduction (in the same issue of Computing Surveys), this survey makes no mention of the data definition facilities in these systems. The paper is at its best when it describes hierarchical data structures and their manipulatio[t, as tools which are useful in the management of data. There is a good tutorial overview of the basic manipulative language, the execution environment, and some implementation considerations. The comparison of two levels of manipulative language is important. Some people judge the merits of a data structure by the kind of manipulative language provided for it; these examples show that different kinds of languages can be implemented for the same structure. The paper Fakes a needless attempt to establish that hierarchies are familiar and natural constructs normally encountered in information. There is an unfortunate play on words here; while we do naturally encounter a number of constructs we call "hierarchical", they rarely correspond to the kinds of structures supported by these systems. In its most widespread use, "hierarchical" refers to things which can be stratified, ranked, classified, etc. But in our use of the term, the restriction to one-to-many relationships is central; I was unable to find this condition mentioned in the definition of "hierarchy" in any of a half dozen dictionaries (both ge, eral and technical). But even in the framework of one-to-many relationships, we are misled. The kinds we encounter very often in nature (and the ones cited as examples in this paperl) are homogeneous trees, which are not the kinds of structures supported by these systems. A homogeneous tree is comprised of relationships among the same kinds of things, as in a personnel organization chart, parse tree, game tree, etc. The structures supported by these systems may only represent relationships between different kinds of things, e.g. , presidents and congresses. Thus, to serve the tutorial function of providin9 a helpful perspective for the …