N-2-3-040.31.1

From Informal Resource Discovery to an Information Infrastructure
Michael F. Schwartz*
<schwartz@latour.cs.colorado.edu>

The U.S. National Information Infrastructure (NII) Act of 1993 calls
for broad deployment of information applications in education, health
care, libraries, and government data.  This represents an enormous call
to action, requiring development of the applications software and the
physical network, as well as efforts to bring historically isolated
agencies and information pools onto the Internet.

Where do current Internet-based information systems fit into this
picture?  Systems like Gopher, WAIS, and WWW have gained loyal
followings, and are widely heralded as important new Internet
applications.  They have caught the attention and imagination of
politicians and the media, and offer an immediate means of bringing some
of the data targeted by the NII online.  However, significant advances
are needed both in the information systems and in the underlying
communications substrate before a true "information infrastructure"
can exist.  I'll start with the systems problems.

First, the information provided by current resource discovery systems is
mostly informal, with no guarantees about data validity or timeliness.
This is not surprising, given the "grass-roots" manner in which most of
the data are placed in these systems.  To some extent, this problem can
be corrected by formalizing (and funding) a support function.  But there
are technical problems as well.  Most of the data have no associated
attributes recording age, seal of authenticity, or even record of
origin.  The systems simply lack the ability to record or act on this
information at present.

Second, current systems provide little or no information structure that
could support a wide range of automated data processing tasks.  The vast
majority of information provided by these systems is intended for human
consumption - formatted text or typeset graphics.  Often the data were
derived from more structured sources (such as a relational database or
spreadsheet), but the disseminated data lack structure.  More
importantly, the current generation of Internet tools provide no means
of representing or accessing this structure.  Mostly they provide flat
keyword searching and menu-browsing interfaces.

Third, the current systems will not scale well for use in a markedly
expanded Internet.  Already it is difficult to navigate around a large
Gopher menu, or choose appropriate keywords when searching a large WAIS
collection.  Imagine what will happen when there is one million times as
much information available.  Other scalability problems exist as well,
most notably in the demand placed on servers of widely popular data.

A number of research and development groups are working on these systems
problems, in both the Internet Engineering Task Force and the Internet
Research Task Force.  The challenge will be to address these needs while
retaining the practical ease of installation and use of the current
Internet discovery tools.  In particular, while X.500 addresses many of
these problems, at present it still requires a good deal of
administrative and computational resources to install an X.500 server.
X.500 servers are deployed at a much slower rate than the other
discovery services.

In addition to the systems problems, work must also be done on the lower
network layers.  Perhaps most importantly, we need support for
commerce-grade (authenticated and private) communications.  Until this
is routinely available, no hospital could possibly risk sending patient
medical history data across the network.  The communications substrate
will also need to support quality of service guarantees, so that, for
example, a medical image being transmitted across the network will not
be delayed by a less important background file transfer.  Of course, an
important component of providing such guarantees is upgrading the
available network bandwidth to meet the increased data demands.

One final problem area is the lack of a workable business model for
developing and deploying information services.  There has been a history
of commercial development and deployment difficulties to date.  For
example, witness the poor responses to "videotext" offerings from the
U.S. regional telephone companies, and to attempts to commercialize the
development of current generation Internet information tools.  With the
beginnings of Internet-based media service offerings (e.g., the creation
of Electronic Newsstand Inc. over Summer 1993), a business model will
hopefully arise that can help lead the way to successful commercial
development of the NII.

*Professor, Computer Science, University of Colorado - Boulder