Edward Yardeni

Leon Kappelman and PhilScott

NIST Federal Processing Standard

Frederick P. Brooks

Corrections or additions?

d

This article by Edward Tenner was published in U.S. 1 Newspaper on

September 29, 1999. Copyright Edward Tenner. All rights reserved.

Y2K — A Crisis of Authority, Not Computers

by Edward Tenner

A former Wilson Center fellow, Edward Tenner is a visitor in the

department of geosciences at Princeton University (E-mail:

tenner@princeton.edu). He is the author of "Tech Speak" and "Why

Things Bite Back: Technology and the Revenge of Unintended

Consequences. An earlier version of this article was published in the

Wilson Quarterly, autumn 1998.

Seventy years ago, W. I. Thomas and Dorothy Swaine

Thomas proclaimed one of sociology’s most influential ideas: "If

men define situations as real, they are real in their

consequences."

Their case in point was a prisoner who attacked people he heard

mumbling

absent-mindedly to themselves. To the deranged inmate, these lip

movements

were curses or insults. No matter that they weren’t; the results were

the same.

The Thomas Theorem, as it is called, now has a corollary. In a

microprocessor-controlled

society, if machines register a disordered state, they are likely

to create it. For example, if an automatic railroad switching system

mistakenly detects another train stalled on the tracks ahead and halts

the engine, there really will be a train stalled on the tracks.

Today, the corollary threatens billions of lines of computer code

and millions of pieces of hardware. Because they were written with

years encoded as two digits (treating 1999 as 99), many of world’s

software programs and microchips will treat January 1, 2000, as the

first day of the year 1900. Like the insane convict, they will act

on an absurd inference. For purposes of payment, a person with a

negative

age may cease to exist. An elevator or an automobile engine judged

by an embedded microprocessor to be overdue for inspection may be

shut down. All of our vital technological and social systems are

vulnerable

to crippling errors. Correcting programs has required time-consuming

close inspection by skilled programmers, custom solutions for

virtually

every computer system, and arduous testing — and time is running

out.

Nobody denies the hazards. And as we will see, if only because of

the original Thomas Theorem, the Year 2000 (Y2K) Problem is already

upon us. The unsettling question is just how serious it will remain

after more billions of dollars are spent between now and then

correcting

and testing affected systems — fully 1,898 in the U.S. Department

of Defense alone, and hundreds of thousands of smaller computer

networks

if those of small businesses are included.

Will the first days of the year 2000 be just a spike in the already

substantial baseline of system failures recorded in professional

forums

such as the Risks site on the Internet? That might be called the fine

mess scenario. Or will it be a chain reaction of self-amplifying

failures

— the deluge scenario? Y2K is not the first serious global

technical

problem, but it’s the first truly global one affecting many different

systems — especially mainframe computers that had never been

infected

by viruses. What’s exceptional about it is contagion. It exposes the

weaknesses of the growing national and global interdependence that

had been points of pride.

The problem also coincides with two other major events: the economic

crisis in Asia and elsewhere, which probably has slowed Year 2000

work in many countries despite partial recovery; and the introduction

of the euro in the European Community, a change that has competed

with already scarce programmers’ time.

Top Of Page
Edward Yardeni

Because Y2Kology mixes evangelism, prophecy, and entrepreneurship,

its message has not won easy acceptance. Read closely, Y2Kologists

share no consensus on how severe the Y2K dislocations are likely to

be. As of February, 1999, Edward Yardeni, chief economist of the

Deutsche

Morgan Grenfell investment bank, estimated the odds of a recession

at 70 percent: 25 percent for a "modest" one, 40 percent for

a "major" one, and 5 percent for a "depression." But

an acknowledged aim of alarming predictions, as in George Orwell’s

"1984," is to galvanize people into action that will prevent

the worst.

By September of this year, most experts agreed that crucial utilities

and financial systems were close to full compliance, though they could

not rule out cascading effects from remaining problems. (Some

opponents

of nuclear power insist that there are still unprepared nuclear

plants.)

Covering all contingencies, Edward Yourdon and his daughter Jennifer

Yourdon have written a guide for coping with a variety of plausible

scenarios, which in their view range from a two-to-three-day

disruption

to a 10-year depression.

And a few panicky Y2K programmers with no evident survivalist

background

have retreated to the western deserts — the very area of the

country

most dependent on electronically controlled federal water distribution

systems.

One thing is certain: the apprehension is real, and will have real

consequences. Just as the fear of nuclear war and terrorism has

transformed

the world over the last two generations, so the mere possibility of

massive system failure will cast a shadow over its political,

military,

business, and scientific rulers for years to come. Year 2000 is less

a crisis of technology than a crisis of authority.

For at least a century, the West has expected, and received, orderly

technological transitions. Our vital systems have grown faster, safer,

and more flexible due largely to cooperation among engineers, state

legislators, and industries to establish uniform codes and inspection

procedures in place of patchwork regulations and spotty supervision.

Most consumers pay little attention to the hundreds of national and

international standards-setting bodies. Only when major commercial

interests are at stake, as when specifications are established for

high-definition television or for sound and video recording, do the

news media report on debates. Laypeople are rarely present at

standards-setting

deliberations.

Before the early 1980s, many conventions were handled

mainly as internal corporate matters. AT&T established exchange

numbers

and area codes, and IBM and a handful of other manufacturers upgraded

operating systems of their mainframe computers.

And why should people worry? The record of these organizations was

unmatched in the world. A Henry Dreyfuss-designed, Western

Electric-manufactured

rotary telephone could work for a generation without repair. Railroads

long ago arrived at standards for compatible air brake systems that

allowed passenger and freight cars to be safely interchanged. And

evolving engineering standards have helped reduce accident levels

on the nation’s interstate highways. The future seemed to be in good

hands.

But no comparable effort has been made to cope with the Y2K problem.

Why?

The breakup of AT&T, the explosion of utilities competition, the

globalization

of manufacturing, and the rise of personal computing have all helped

diffuse authority over standards. And freedom from regulatory

entanglement

has brought immense benefits to manufacturers, consumers, and the

economy. But it has had an unintended consequence. The diversity of

systems and the fierceness of business rivalries discourage public

and private technological authorities — from the Defense

Department

to Microsoft — from taking firm and early action to cope with

emerging problems. In general, governments have avoided interference

in commercial decisions, and businesses have succeeded more by

following

market shifts than by staking out ambitious new standards. As the

Thomas Theorem implies, if people do not believe they can exert

influence,

then they cannot. Which brings us to "the millennium bug,"

which is no bug at all.

Over the last four decades, the Year 2000 Problem has passed through

three phases, each bringing its own challenges for authorities. The

first age, the Time of Constraint, lasted from the origins of

electronic

computing to the early 1980s. The managers and programmers of the

time knew that programs using only two-digit years had limits, but

there was a strong economic case for two.

Top Of Page
Leon Kappelman and PhilScott

Leon Kappelman and the consultant Phil Scott have pointed out that

the high price of memory in the decades before personal computing

made early compliance a poor choice. In the early days of computing,

memory was luxury real estate. A megabyte of mainframe hard disk

storage

(usually rented) cost $36 a month in 1972, as compared with 10 cents

in 1996. For typical business applications, using four digits for

dates would have raised storage costs by only one percent, but the

cumulative costs would have been enormous. Kappelman and Scott

calculate

that the two-digit approach saved business at least $16-$24 million

(in 1995 dollars) for every 1,000 megabytes of storage it used between

1973 and ’92. The total savings are impossible to calculate, but they

surely dwarf most estimated costs of correcting the Year 2000 problem.

(One leading research group, the International Data Corporation,

estimates

a correction cost of $122 billion out of more than $2 trillion in

total information technology spending in the six years from 1995

through

2000.)

Even where Year 2000 compliance was feasible and economical, it wasn’t

always in demand. In the 1980s, a number of applications programs

were available with four-digit dates, such as the statistical programs

and other software systems produced by the SAS Institute, one of

computing’s

most respected corporations. SAS does not appear to have promoted

it competitively as a major feature. The Unix operating system,

originally

developed at Bell Laboratories, does not face a rollover problem until

2038, yet this too did not seem to be a selling point. Even Apple

Computer did not promote its delayed rollover date of 2019. The year

2000 still seemed too far away.

By the mid-1980s, the Time of Choice was beginning. The economic

balance

— initially higher storage and processing costs versus long-term

savings in possible century-end conversion costs — would have

still been an open question, had it been openly raised. The great

majority of crucial government and business applications were still

running on mainframe computers and facing memory shortages. But the

trend to cheaper memory was unmistakable. The introduction of the

IBM PC XT in 1983, with up to 640 kilobytes of random access memory

(RAM) and its then-vast fixed hard drive of 10 megabytes, was already

signaling a new age in information processing.

Yet the possibilities presented by the new age remained an abstraction

to most computer systems managers and corporate and government

executives.

Then as now, most of their software expenses went not to create new

code but to repair, enhance, and expand existing custom programs —

what are now called "legacy systems." A date change standard

would initially increase errors, delay vital projects, and above all

inflate budgets. And it was not a propitious time to face this kind

of long-term problem.

The American industrial and commercial landscape during the 1980s

was in the midst of a painful transformation, and investors appeared

to regard most management teams as only as good as their last

quarter’s

results. Only the mortgage industry, working as it did on 30-year

cycles, had recognized the problem (in the 1970s) and begun to work

on it.

Top Of Page
NIST Federal Processing Standard

Not that government was much more prescient. The Federal Information

Processing Standard of the National Institute of Standards and

Technology

(NIST) for interchange of information among units of the federal

government

specified a six-digit (YYMMDD) format in 1968 and did not fully change

to an eight-digit (YYYYMMDD) format until 1996. The Social Security

Administration was the first major agency to begin Year 2000

conversion,

in 1990. The U.S. Air Force used single-digit dates in some 1970s

programs and had to have them rewritten in 1979. Despite the

impressive

military budget increases of the 1980s and the Pentagon’s tradition

of meticulous technical specifications for hardware, many vital

Defense

Department systems still require extensive work today.

The computing world of the 1990s recalls a multimedia trade show

display

decorated at great expense and stocked with the best equipment money

can buy, yet still dependent on a hideous, half-concealed tangle of

cables and power lines, with chunky transformer blocks jutting

awkwardly

from maxed-out surge protectors. Our apparently seamless electronic

systems turn out to be patched together from old and new code in a

variety of programming languages of different vintages. The original

source code has not always survived. Year 2000 projects can turn into

organizational archaeology and confront us with many such example

of engineering coexistence.

During the Time of Choice, the problem was recognized

but deferred for two reasons. First, there was the chance that entire

computer systems would be replaced before 2000. Second, future

software

tools might reduce conversion costs sharply. Y2K work always has had

opportunity costs. There has always been something that seemed either

more urgent, more profitable, or at least "cooler." (Even

now, I have heard of no Y2K tycoons. Some Y2K consultants have

diversified

successfully into other computer advice, but the stock of the public

specialized Y2K firms as a group has declined sharply.) Furthermore,

turnover in executive ranks and expectations of earnings growth,

discouraged

long-term thinking at the very time that it was called for — in

the mid to late 1980s.

The Time of Choice ended in the early 1990s, when leading computer

industry publications prominently recognized Year 2000 conversion

as a problem and warned of the consequences of neglecting it. It was

followed by the Time of Trial in the mid-1990s, as conversion programs

began in earnest and Y2K issues were increasingly aired in the

computer

press. It will probably last until around 2005.

A few annoyances are already apparent. Credit cards with 2000

expiration

dates, for example, have been rejected by some authorization systems.

Y2K optimists are taking heart at the absence of major problems at

some anticipated trouble spots: the beginning of fiscal year 2000

in many organizations over the summer, and more recently the rollover

of older global positioning system (GPS) equipment and the date

9-9-99,

once used by programmers as an arbitrary date.

Top Of Page
Frederick P. Brooks

If the coexistence of past, present, and future was the discovery

of the Time of Choice, triage is becoming the watchword of the Time

of Trial. Fortunately, information technologies are not created equal.

Some organizations have hundreds or even thousands of computer

systems,

but only a minority, are vital and only a few may be critical. It

is too late to fix everything, even with emergency budgets and the

mobilization of computer-skilled employees from other departments.

As the project management guru Frederick P. Brooks pointed out in

his classic "Mythical Man Month" (1982), adding programmers

to a late project can actually delay it further. In a complex

interconnected

system, more things can go wrong.

In the Time of Trial, triage will not be the only military metaphor.

Many other information technology projects will be suspended or

canceled

as programmers are called up for the front. Careers will be damaged

and entire organizations will be set back. Well-prepared companies

will gain strategic advantages.

Expert opinion now expects a fine mess rather than a deluge. Banks

and investment houses have been advertising their Y2K compliance.

Organizations that feared they would not make the deadline have

installed

new hardware or software or both, outsourced vital functions, or even

merged into larger companies to avoid paralysis. These decisions have

brought problems of their own, like the difficulties of some

universities

running the systems of the human resources vendor PeopleSoft, but

the new glitches are not attributed to date changes. (It is likely,

in fact, that the biggest impact of the Year 2000 episode will not

be date-dependent problems that were missed, but unrelated errors

introduced in the myriad lines of code introduced or changed for Y2K

compliance.)

But some dangers will persist despite the efforts of even the most

resourceful managers. Realization of any one of the five most ominous

threats could validate the doomsayers’ predictions. These risks might

be abbreviated as SMILE: second-order effects, malicious code,

interdependencies,

litigation, and embedded processors.

Thomas’s Theorem suggests that the expectation of a Year 2000 crisis

may be enough to create a real one no matter how effective the efforts

to repair the underlying code. Our social and technological systems

are more efficient than ever, but because, for example, information

technologies now allow vendors and manufacturers to maintain lean

warehouse inventories, slight disruptions can have more serious

repercussions.

Running the gamut from shifts of investment funds based on

Internet-transmitted

rumors about Y2K readiness of particular companies, to depletion of

bank and automatic teller machine currency supplies, to runs on bread

and toilet paper, a late 1999 panic might be comical but also

potentially

deadly.

Add potential sabotage to the equation. The Pentagon already worries

about information warfare and terrorism. Hostile states, criminal

organizations, and domestic and foreign radical movements can already

attack vital networks. The beginning of the year 2000 is a perfect

cover. Do not forget embezzlers and vengeful staff. An apparently

Year 2000-related incident could mask electronic robbery, and a

continuing

shortage of skilled personnel could delay diagnoses for priceless

months. Computer security experts also fear fly-by-night Y2K

consultants

who may collude with corrupt managers to offer bogus certification,

or plant Trojan horse programs in the systems of honest but desperate

ones.

Thanks to decades of global thinking, North America

and Europe are also linked to nations whose Year 2000 readiness makes

many Western nations look like paragons. The Asian financial crisis

that began in 1998 has surely delayed the compliance programs of some

major trading partners of the United States and Europe. International

interchange of data may send a failure in one country rippling through

the most rigorously Year 2000-ready systems: the sociologist Charles

Perrow calls this "tight coupling." Major corporations are

already pressing their trading partners for certification of their

Year 2000 compliance. Domestically, this may make or break some firms,

but it will not bring down the economy. Internationally, it may

trigger

local crises that might lead to mass migrations or insurrections.

And even if all suppliers can be certified, who will verify compliance

of their subcontractors? In fact, any attempt to make Y2K

certification

universal, down to the last level of sub-vendors, would trigger

massive

correspondence likely to disrupt commerce as much as any actual

software

failure.

The courts have only begun to consider legal liability for Year 2000

failures. The cases already on the docket will test one of the law’s

principles: to decree retroactively but to create predictability.

Because Year 2000 cases will raise new questions and provoke immense

claims, the litigation will be prolonged and possibly ruinous. On

the other hand, recent federal legislation limiting Y2K liability

might prevent businesses and individuals from collecting justifiable

claims, according to consumer advocates and some Y2K experts. Just

because product defects are electronic rather than mechanical or

chemical,

should plaintiffs’ rights be curtailed just because many firms were

negligent? It remains to be seen whether legislatures and courts can

deflect nuisance suits while protecting users.

The most serious wild card of all, though, is a hardware issue. Most

discussions of the Year 2000 Problem focus on the difficulty of

repairing

and testing software, but that is a cinch compared to dealing with

the thousands of embedded microchips that control critical systems.

The Gartner Group estimates that 50 million embedded devices may

malfunction.

Traffic signals and freeway entrance metering lights will fail.

Elevators

will shut down if their electronic hardware tells them they have not

been inspected for nearly a hundred years. (The largest elevator

manufacturers

deny their products are vulnerable to Y2K failure.) Electric power

distribution switches and pipeline controls will interrupt energy

flow. X-ray machines will not turn on — or far worse, off —

at the proper times.

The Year 2000 Problem shows that neither military nor civilian

authority,

neither social democracies nor authoritarian regimes nor market

economies,

neither big business nor small business, took fully adequate steps

in planning for the future. If centralized technological planning

is discredited, if the discipline of markets (such as securities

analysts’

reports and insurance underwriters’ risk assessments) has failed to

give timely warning that cannot be ignored, what is left? Perhaps

it is the realization that technology is not just a radiant future

but a messy present, that the age of transition never ends, and that

rapid novelty, massive legacy, and tightly-coupled systems can

interact

to create lethal assumptions.

Fortunately, crises have nearly always stimulated important and

productive

innovations. They call attention to people and ideas outside the

establishment

— in this case, programs in information technology management

as opposed to the older and larger MBA programs and computer science

departments. Tools and concepts emerging from Year 2000 remediation

and testing may well have uses that extend to future conversion

questions.

As usual, it’s hard to say what lessons these will be until we learn

them.

January 1, 2000, will not be the first danger point, and it will be

far from the last. The outcome of Y2K will change everything, but

if we already knew what will be changed, there would have been no

Year 2000 crisis, only a problem. Making systematic correction and

recovery easier may be the hardest job of all. But our pace makes

it necessary. As the Red Queen said in "Through the

Looking-Glass:"

"Now, here, you see, it takes all the running you can do, to keep

in the same place. If you want to get somewhere else, you must run

at least twice as fast as that!"


Next Story


Corrections or additions?


This page is published by PrincetonInfo.com

— the web site for U.S. 1 Newspaper in Princeton, New Jersey.

Facebook Comments