Object lens: a “spreadsheet” for cooperative work

Reviewer: Joseph L. Podolsky

The use of objects has become the latest fad in software engineering, picking up where artificial intelligence left off. The description of the Object Lens in this paper is useful, however, because it provides some specific examples of how objects might work in a real-world environment and offers the notion of semiautomated, semiautonomous software behavior. The Object Lens is a user interface that integrates hypertext, object-oriented databases, electronic messaging, and rule-based intelligent agents. Still in the prototype stage, this system succeeds the earlier, less capable Information Lens and is designed to be a knowledge-based environment for developing applications in a cooperative fashion, where the work is distributed among several people, teams, or geographic locations. The authors try to relate their system to commercially successful spreadsheet packages by offering a very general definition of spreadsheet software. The success of products like Lotus 1-2-3, however, was largely due to a factor they do not mention—the target market's familiarity with the paper spreadsheet. Although the Object Lens does incorporate elements that have analogies in nonautomated processes, the authors do not state these analogies explicitly, and the system has no single manual metaphor. It does, however, resemble a meeting where project team members exchange various finished pieces of programs and documentation that other team members then use. The authors call their system “semiformal,” that is, one that represents and automatically processes certain information in formally specified ways, represents and facilitates the human processing of the same or different information in ways that are not formally specified, and allows the boundary between formal processing by the system and informal processing by humans to be changed easily and frequently. Their system can also represent and manipulate “semistructured” objects; fields can be structured or not, and the data in them can be restricted or flexible in type or content. The Object Lens uses “template-based” user interfaces, and displays may be defined by the user. As in other object-oriented software, all elements of the software can be treated as objects to be called by or embedded into other objects. New objects can inherit some or all of the characteristics of the old objects, and objects can be collected into folders and linked into tables or tree structures. Semiautonomous agents can process information. These agents can take a series of actions without human intervention; the processing rules, however, are easily visible to human users, who can change them. The agents may also refer objects to a human user for action. Agents are also objects, of course, that can be called by or embedded in other objects with full or partial inheritance. Frankly, this paper did not add much to my knowledge of objects, but the notion of human/computer interaction that the various “semi-” functions offer did impress me. I know too much about computers and software to trust them very much, and I like the idea that the power of this system is visible to and under the convenient and explicit control of the human user. The Object Lens is currently a prototype implemented in Interlisp-D on Xerox 1100 workstations. The authors claim that all the features they describe are actually implemented, although some are not fully tested. This good work in progress may eventually result in a powerful family of products, but even if the Object Lens should fail in the marketplace, the ideas in this paper are valuable and will further research in the area by ratcheting our expectations a few notches higher.