Principia Cybernetica Web has a tree-like structure: each node, except the Home node, has usually one, sometimes more, parent nodes, which are hierarchically superior to it, the way a section of a book is on a higher level than a subsection. In turn, the node has normally several child nodes, which are hierarchically inferior.
If there were always just one parent node, the structure would be a pure tree. However such a structure is rather rigid, since it implies that every subject is unambiguously classified as part of a single more general subject. This is a recurrent problem in subject indices, as they are used, for example, in libraries. Should a book on biophysics be classified under "Biology" or under "Physics"? The problem is tackled in PCP Web by allowing multiple classification. For example, the node entitled "Progressive formalization" describes PCP's view of the process of reformulation of scientific models in order to make them more precise. It fits in as well with the theoretical section on epistemology, as with the practical section describing the methods used in the development of PCP's knowledge base. Therefore, the node has two parent nodes.
Still, a tree-like ordering or classification has great advantages in reducing the complexity of the system, since it minimizes the number of paths that lead from one node to another. Moreover, there are theoretical reasons for assuming that hierarchical orderings arise naturally, and therefore provide in general a good description of most systems (see the Principle of Recursive Systems Construction). Therefore, the default is one parent node, and the nodes classified under two (or more) headings remain the exceptions. We call such a structure a loose or "quasi-" hierarchy. Practically, it means that for most nodes there is a single path from the "top" node of the hierarchy (the "Home page" described earlier) via the different children and grand-children down to the node in question.
This quasi-hierarchical ordering is reflected in the Table of Contents: child nodes are distinguished from their parent by one level of right indentation, as is common in "outliner" software. Nodes at the same level have the same indentation. Like in outliners, the "multiple parenting" is represented by simply repeating the description of a node under each of its parents (this is sometimes called "cloning" in outliners).
Of course, the selection of parent or child links is not the only way to navigate from node to node: this would defeat the flexibility characteristic of the hypertext concept. An unlimited number of links to other nodes can be present within the main text of a node, introduced informally by phrases such as "see also" or simply by highlighting a reference or a technical concept, with an included link to the place where it is defined. The parent-child hierarchy, represented by the separate fields, functions as a kind of "skeleton" within the overall, free-form web of links between nodes, allowing an unambiguous localization of each node as a specific subsection of the network.
In addition to the hierarchical ordering, the tree structure entails a linear, sequential ordering of the nodes (listing order in theTable of Contents). This can be interpreted as a suggested reading sequence, (implemented through the "Next" command in the menu bar of each node). Again, this order is merely a kind of default structure, which can be freely ignored. However, for readers wishing to systematically study the corpus of knowledge in Principia Cybernetica Web, the following of this linear path guarantees an exhaustive covering of the material. Moreover, such a path makes it possible to convert the hypertext web to a conventional linear text, which could be printed out as a book. Note that it is possible to create personalized paths, covering different parts in different orders, depending on the preferences of the reader or author.
