a 10 (TEN) year old document that still seems up to date

[David Farber <farber () central cis upenn edu>]

This document was prepared and presented approximately 10
years ago. We have prgressed so relately little that I
could just re-date it and resubmit it. I thought you might
want to reflect on this


Dave






Information Systems Engineering Perspectives


David J. Farber
University of Pennsylvania Department of Computer and Information Science
Philadelphia, Pa 19104-6389


Phone:  (215) 898-9508
Internet:  farber () cis upenn edu


15 July 1985


Abstract


This document presents an overview of the state of the information systems
engineering research area. It is found that the situation in the academic world
is far from ideal. A number of areas of research where the nation and the field
could benefit from increased attention are discussed; and equally importantly, a
set of needs is highlighted, which, if satisfied, would
increase the productivity of the academic researcher.
Finally, a set of recommendations will be made for specific
activities that could lead to the creation of a new
generation of information systems engineers.


Prepared for the Policy Research and Analysis Section of the NSF, and presented
at the Internal Workshop on opportunities for Engineering Research Focused on
Emerging Engineering Systems July 1985. Introduction


This report will examine the future directions for information systems
engineering at the National Science Foundation. It will address the current
status of the field, its current problems, its importance to the nation, and
potential directions for enhancing the impact of the NSF in this area. In
addition, it will examine a set of alternative scenarios that the NSF could
follow and the impact of each on the future of the field. Finally, it will
propose several potential areas for research concentration that could have high
payoff for the nation.


Background


There have been many studies of the information sciences and technology areas
done over the past several years. Perhaps the most
comprehensive of these studies was one done in 1984 by the
Office of  Technology Assessment, Congress of the United
States, entitled Information Technology R & D:  Critical
Trends and Issues. A  copy of the summary document has been
attached to this report as
Appendix A. It would be appropriate to examine the principle findings of this
report (printed in italics)  and comment on each.


Principle Findings of the OTA Report


Most areas of information technology examined in this study, including
microelectronics, fiber optics, artificial intelligence, computer design, and
software engineering, are still in the early states as technologies. In
particular, in the software engineering area, I and others have pointed out the
lack of coupling of university research activities to the
commercial environment,
where large software systems are written. Later on in this document, I will
specifically comment about this and other lapses of technology transfer.


By most measures, U. S. research and development in information technology is
strong and viable; however, those traditional measures may not be realistic
guides to the future needs of the United States for R & D in these areas.


In response to these new pressures (foreign competition and profit motivations),
industrial support is growing rapidly for short term  applied research and
development work, both within industrial labs and through support of university
work.  While industry has traditionally looked to the academic world for basic
research support in many other areas of science, it has over the past decade
ceased to expect such in the information systems engineering area. Many of the
joint academic-industrial programs have resulted in the university becoming a
development shop for industrial researchers. In addition, the quality and
quantity of academics who can work with and understand the industrial sector has
decreased in this field due to the large number of entrepreneurial start-ups
motivated and manned by ex-academics. We are finding faculty positions in
academia increasingly filled by people who have inadequate experience to
understand, react, and deal with the real problems of industrial research and
advanced development.


Universities, traditionally viewed as centers for basic research, are
re-examining their roles with respect to applied research and are forming new
types of relationships with industry and government. The OTA report states that
it is too early to say whether or not this will have a negative or positive
response on university. It is my opinion that this has
already been shown to have
a negative effect on the underlying academic role of the university. What has
happened in a large number of cases is that the institutions that have been
formed are decoupled from the university, thus exacerbating the isolation of the
teaching program and graduate research from the institutional activities. The
University has effectively created a corporation for doing external research
which is indistinguishable from a separate R & D entity, usually without gaining
the full benefits accruable from such. This has created
the phenomena of faculty
who never teach and have minimal contact with undergraduates and even with
students, because they spend a majority of their time in these institutes. I
strongly question the advisability of encouraging such directions.


The Department of Defense is the predominant source of federal support of
information technologies research and development, providing nearly 80% of the
funding. Experience has shown that the  spillover from these activities to the
civilian industrial sector is minimal in the information
systems area because the
cost effectiveness criteria for military research and development is totally
different from that in the commercial sector. There are, of course,
counterexamples, such as ARPANET. However, many senior industrial managers
believe that such examples of significant spillover are few and far between, as
witnessed by their reluctance to get involved in some DOD
activities, such as the
very high speed integrated circuit work. There are examples within U.S. industry
where major corporations refuse to undertake DOD research or have isolated this
work into separate subsidiaries. Experience has shown that the technology
transfer from these subsidiaries to the mainline commercial areas is essentially
nonexistent. Both the OTA and I strongly believe that increased funding for
long-term research in information systems and technology is needed from
non-defense agencies in order to focus research on areas that will have more
civilian payoff.


There is substantial concern that technical and scientific information flows
between the U. S. and other countries are unbalanced
outward. These concerns have
recently surfaced in the form of the new Department of Commerce Export Control
Laws. A tight interpretation of these rules would cause a severe decrease in the
quantity of graduate students available in the information systems area, since
many of these students come from abroad. Additionally, the
decrease in freedom to
publish may have a severe impact on our ability to communicate with our peers,
even within the United States. While I do not advocate a completely open
information flow, it must be realized that such constraints will have a
particularly acute impact on the field of information systems engineering.


Instruments for scientific research are growing more sophisticated and are
becoming obsolete at an increasingly rapid pace. This is
particularly apparent in
the area of information systems engineering. The computers
used as research tools
become obsolete at a  rapid rate. Such obsolesence especially causes problems in
this field, since research often relies on advanced hardware and advanced
communications facilities. Outdated equipment puts the academic researcher in a
disadvantageous position relative to his industrial research colleague. This
disadvantage accelerates the departure from the academic
world of the talent that
is necessary to both properly train the next generation of
students and to insure
that active, relevant research is maintained in the academic community. An area
where the lack of adequate equipment is particularly severe is in the computer
architecture and computer communications disciplines (as well as, of course, in
microelectronics). In these cases, academic researchers who are trying to
investigate new computer organizations or new communications systems are
definitely handicapped by the lack of modern fabrication facilities, of state of
the art CAD systems, and of the support staffs which are norms in equivalent
industrial laboratories. Efforts on the part of non-DOD funding agencies to
attack the equipment problem in information systems
engineering have been minimal
as are current efforts to allow the academies effective access to industrial
facilities.


Policies designed to stimulate information technology R & D need to be evaluated
for possible significant tradeoffs and external costs in other areas.


End of OTA summary


What is the State of Information Systems Engineering Research?


Advanced Computer Architecture


While we have been through what appears to the public to have been several
revolutions in computer architecture, little has been done at the fundamental
organization level. The architecture that are now in use in
the commercial world,
with several notable exceptions, were designed over 15 years ago. Even our
technological star, the microcomputer, has an internal
organization which has not
fundamentally changed over the past decade. There are many reasons for this
phenomena; some are commercial in the sense that new and innovative designs have
been very hard to sell in the commercial marketplace (-for example, the Intel
iAPX 432). Safe, well-understood designs which are upwardly compatible with past
generation computers have tended to be the norm.


While such conservatism  is not in and of itself bad, it does have a dampening
impact on the innovation that one can expect from the commercial field.
Innovative architectures is an area where the university, with its studied
indifference to commercial viability can have a major impact. One can argue that
the RISC architecture re-spawned from the university environment has become the
basis for several new and innovative microprocessors (there are also opinions
that RISC is just a transient reaction to technological
tradeoffs). Nevertheless,
the question of how such innovation can be encouraged and
harnessed is one of the
most difficult issues facing the planners in our national research supporting
agencies.


I believe that it is critical to support the academic engineering research
community in the computer architecture area. Such support must lead to the
creation  of an infrastructure which will allow researchers to try their ideas
and to create new designs in a reasonable amount of time. A path towards
attaining this capability could be modeled on the solution to a similar problem
that the NSF faced 20 years ago with the advent of computers as a research tool.
At that time, the NSF undertook a program of hardware capability grants to the
universities on a massive scale to to seed the computer science programs that
were forming at that time. A similar program, centered about providing
state-of-the-art CAD tools, wide spread access to silicon foundries on a rapid
turn-around basis, and  modern architecture verification and simulation tools
running on state-of-the-art engineering workstations would
have a profound impact
on the ability of researchers to operate in the university environment.


I strongly recommend that a major initiative be undertaken to supply to
university departments in the information systems engineering area,
state-of-the-art CAD facilities and widespread access to fast-turnaround
fabrication facilities. These fabrication facilities should include both
board-level and microchip-level capabilities. In addition, a mechanism must be
found to fund adequate technical support staffs so that the maximum productivity
can be achieved by researchers.


Software Engineering


The field of software engineering has received a significant amount of attention
over the past 5 to 10 years as the balance of effort and cost in the development
of new systems shifted primarily to software development.
Many computer companies
(especially microcomputer companies) have found, much to
their horror, that their
software staffs are 2 to 3 times larger than their hardware staffs.


Very large, defense-oriented software activities have
pointed out the sorry state
of our knowledge of how to write large software systems. The ability to create
relatively  free bug software at a tolerable cost has
become perhaps the deciding
factor in the battle for dominance of the computer industry. It is one of the
areas where the United States has shown itself to be traditionally stronger than
our foreign competitors, but is again under attack by Japan and Europe.


It goes without saying (however, I will say it), that the feasibility of large
weapon systems, such as SDI, depends on strong software engineering technology.
Likewise, the widespread success of large, distributed applications, such as a
fund-transfer systems, automated factory systems, real-time control systems,
etc., depends on low-cost, high-reliability software systems.


To date, much software engineering research has centered on the theoretical
aspects of software. While such studies may provide the basis for long-term
payoffs, they have yielded little insight that can be used in the short run by
practicing software designers. We have not yet achieved the fundamental
breakthroughs that will be necessary to achieve real success in this field. In
the meantime, the commercial field is suffering from excessively expensive, ever
more fragile software systems.


Perhaps a major reason for this lack of short-term benefits
for the field lies in
the lack of exposure of academic researchers to large-scale
software development.
Those who do develop skills in this area leave academia to form private
companies. The problem of how to involve academics in large-scale activities so
that they might learn the problems, and thus contribute the solutions, appears
insurmountable. The only path that seems at all viable lies with
industrial/academic collaboration at a level which has not
yet been practical. It
would involve a close liason between the researchers and ongoing industrial