Roy T. Fielding,
Maintaining Distributed Hypertext Infostructures:
Welcome to MOMspider's Web.
All documents on the World-Wide Web can be considered part of an infostructure -- an information resource collection with a specifically designed structure [Tilton93]. An infostructure is created any time an amount of information is organized in a useful manner, such as a dictionary or this paper (Figure 1). In the WWW, infostructures can include a wide variety of information sources, in the form of interlinked documents, and each document may be shared by multiple infostructures. In turn, an infostructure can be contained within higher-level infostructures, just as the table of contents is contained within the overall structure of the hypertext version of this paper. The World-Wide Web as a whole can be considered the ultimate online infostructure.
A good infostructure doesn't just exist; it exists by design. The documents are linked specifically to allow readers to navigate through the information presented. If that information is presented as hypertext, the reader should not be constrained by the linear nature of traditional documents. Therefore, hypertext authors attempt to provide as many routes of navigation as are conceived to be useful for potential readers. Each hypertext link creates a dependency between the source and destination documents.
In a world of information, most of the best sources are maintained at sites other than the author of a particular infostructure. Furthermore, such information is rarely static, consisting of living documents maintained by the owner (usually, but not necessarily, the original document author) at those distributed sites. Rather than copy the contents, the WWW enables infostructures to be composed simply by referencing the desired object's URL within the guiding text of any HTML document. In addition to providing document reuse, this allows new routes of navigation to be developed by the consumers of that information, independent of the design considerations of existing infostructures.
Unfortunately, the flexible and dynamic nature of the World-Wide Web leads to a glaring problem -- one that should already be familiar to many information providers. Most infostructures are time dependent. They reflect the navigational desires and information contents that were available at the time of their design. However, as living documents change and new documents are added, the resulting structure may vary from that intended. Documents may be moved or deleted, referenced information may change, and hypertext links may become broken. A living infostructure must therefore be actively maintained in order to prevent structural collapse.
The maintenance problem is exacerbated by that of scale. As it grows, an infostructure becomes complex and difficult to maintain. Each new document may result in an exponential increase in document links. A large infostructure will therefore require many human maintainers, particularly if it is spread over multiple distributed sites. However, this also compounds the problem -- as individual maintainers make changes to support their own part of the infostructure, those same changes may break the structures maintained by others.
[Continue to Limitations of Existing Maintenance Methods or Up to Contents]