son, he and his team developed NLS, the first working hypertext system with graphical user interface, mouse, and collaboration tools.

In the late 1980s, Tim Berners-Lee created the World Wide Web infrastructure to facilitate information access and as a potential means to implement many of Bush’s, Nelson’s, and Engelbart’s ideas in the established worldwide Internet.

While these systems were major advances in the quest for universal information access, they left unsolved many of the difficult issues of information management noted earlier. Our objective here is to show that the Digital Object Architecture, which has been under development since the late 1980s at the Corporation for National Research Initiatives (CNRI) and is now reaching a tipping point, can provide the missing pieces of infrastructure and help us to attain universal information access.

 

Toward a universal address Space

A universal address space is the most fundamental element of a system of universal information access. A collection of ARPA-sponsored projects in the 1960s developed the first methods for doing this efficiently.

In a famous 1960 article “ Man-computer symbiosis,” J.C.R. Licklider expounded on the virtues of “ intergalactic networks” and man-machine communications. Partly at his instigation, MIT’s Project MAC undertook the construction of an operating system, Multics, which would be a “computer utility” that could dispense computing power and information access widely and cheaply. Other research organizations, such as IBM, UC Berkeley (Genie), and BBN (Tenex), were early developers of time-shared operating systems that could be networked.

Within the Multics system, information sharing was achieved by making virtual memory a common space accessible by every user on the system. The directory structure was an overlay that let users assign their own symbolic names to files; files themselves were addressed internally by their virtual addresses. Users never had to open or close files, or copy them from secondary storage to their workspaces; they simply referenced them as segments of address space.

a universal address
space is the most
fundamental element
of a system of
universal information
access.

Jack Dennis, who helped design the Multics virtual memory, saw how to generalize it to allow dynamic sharing of a more general class of objects, not just files, and how to protect these objects from access not permitted by their owners. His “capability architecture” became the blueprint for object-oriented runtime systems. 2 That architecture inspired two commercial computing systems—Plessey 250 and IBM System 38—and two research projects— Cambridge CAP and CMU Hydra—that contributed much implementation knowledge. These projects all demonstrated that a large widely accessible address space would not only facilitate sharing, but it could be implemented efficiently on a single machine with a large shared file system. The capability architecture also provided a clean way of managing access and controlling rights by channeling all object references through a reference monitor protocol. 3

The capability architecture easily generalized to homogeneous networks all running the same operating system. Unfortunately, it did not generalize well to the heterogeneous systems making up the Internet.

In 1970, the ARPANET expanded the universe of connected systems beyond a single machine to hundreds of time-shared computers. In the U. S. and around the world, other packet networks were developed in parallel with the ARPANET, including, in particular, a packet satellite network, a ground radio packet network, and various local area networks, such as token rings and Ethernet.

The Internet, introduced in 1973 by
Bob Kahn and Vint Cerf, is a global in-
formation system composed of many
networks and computational resourc-
es all embedded within a single large
address space based on the Internet
Protocol (IP) addresses. The domain
name system, introduced in 1983,
maps domain names to IP addresses,
making it much easier for users to ad-
dress computers.

 

The Digital object architecture Universal information access was not achieved in the Internet because none of the protocols was designed relative to information-sharing architecture principles. In the 1980s, prior to the development of the Web, as part of its work on digital libraries, CNRI started designing a system for enabling mobile programs (called “Knowbots”) to carry out information-access tasks in a network environment. This led to the later formulation by CNRI of the Digital Object Architecture (DOA), 1, 4, 5 which culled out and unified four key