SMART Letter #6:
Unknown Year 2000 Unknowns
May 10, 1998

               SMART Letter #6 - May 10, 1998
        For Friends and Enemies of the Stupid Network
            Copyright 1998 by David S. Isenberg
      This document may be redistributed provided that
      the 11 lines containing this notice accompany it. -- -- 1-888-isen-com
      It takes SMART people to design a Stupid Network


The Viagra economy of 1998 makes it easy to ignore the Year 2000
Problem.  Yet there is reason to surmise that we have passed the
point of no return -- that we are likely to have serious systems
failures as we transition to the Year 2000.
How serious?  We don't know.

Some of these are likely to be of the expected,
first order variety -- bank screw-ups, supply chain problems,
air traffic control delays, etc. -- but we might also experience
emergent, higher-order unravelings, with consequences that could
be economic, social, medical, geopolitical.

On the other hand, we might *not* experience such emergent,
society-shaking consequences -- and I *hope* not -- but contingency
thinking today beats being surprised and unprepared tomorrow.

Complex adaptive systems - such as ecologies or markets - rarely
progress in a smooth linear fashion for long.  The space folds.  If there
are too many rabbits, suddenly coyotes appear.  Virulent strains of
disease course through populations, decimating their host until resistant
individuals meet weaker strains of disease to establish a new dynamic
balance.  Markets feed on their own enthusiasm, bubble and burst, then
"correct" and regain temporary, perhaps illusory stability.

Yet we have progressed, apparently smoothly, from a world of isolated
8 kilobyte machines to a gigalink society, a network of pervasively
networked networks, whose methods and meanings are emerging faster
than we can discover them - let alone understand their implications.

Moore's Law was a wild card.  Nobody planned how the individually
hand-crafted computer programs of the 1970s would acrete to form the
Just-in-Time Economy that supports today's industrialized world.
We who wrote Fortran in that era are surprised that our morsels of
kludgey code, painstakingly crafted to use expensive memory with utmost
efficiency, would still be running today.


The Year 2000 Problem is born of well understood, widely accepted, once
omnipresent programming practices that were established when memory was
expensive.  (Remember how we named a variable D instead of DAY, or S
rather than SUM, to save memory?  I do.)

Memory got cheap, but the code we wrote is still running.
Thus, the Y2K problem is widely sown in older code.
Often it appears in mutated, hard to recognize patterns.  Or in
systems that are three miles under the ocean or 300 miles out in space.


It is easy to overlook, ignore, and minimize the disruptive potential
of the Y2K Problem.  I was reminded how easy Y2K is to deny when the
May 4, 1998 issue of the new, otherwise technologically savvy webzine,
The Industry Standard ( scoffed that
Y2K was being over-hyped by greedy, scare-mongering consultants.

Why so easy?  A few reasons.  In isolation, it is supremely boring
whether the year is a 2-digit or 4-digit field.  And nobody likes to
dwell on bad news, especially in good times.  And facts about Y2K, when
you can find them, are either boring and technical, or overly dramatized
and, indeed, over-hyped.  And good observations on the emergent, systemic
nature of the problem are difficult to find.  And sometimes known facts
are actively suppressed by good people who fear litigation.

So here are some observations:
People who have not been involved in Y2K assessment are prone to say,
"No problem."  In contrast, the people with the longest faces, who
speak of Y2K in the gravest tones, belong to the firms that are
furthest along in the remediation process.  Over the last year,
four out of five corporate Y2K budgets have been increased upon re-
evaluation.  As FCC Chairman William Kennard said recently, "Every company,
every government agency, and every organization that has looked into the
problem has found that it is more complicated, more serious and more costly
than originally estimated." (April 28, 1998)

data, collected over 30 years by software project expert Capers Jones,
shows that as the size of a software project increases, the more likely
it is to be delayed or cancelled.  Thus, a 100 line project has an 92%
probability of being completed on time, a 10,000 line project has a 62%
chance of on time completion, and a 1,000,000 line project has a mere
14% chance of on time completion (and a 24% chance of outright

Repeat:  A one million line project has a 14% chance of on time
completion.  The FAA has 23 million lines of code, Prudential Insurance
has 160 million lines, Chase Manhattan Bank has 200 million, Citibank
has 400 million, and AT&T reportedly has 500 million.

The farmer's fertilizer is manufactured under computer
control.  It is delivered by a truck that is dispatched by an information
system.  The planting and harvest are guided by information systems.
The food is processed and shipped under the control of information
systems.  The feed and seed store, the farmer, and the middleman are
paid by information systems.  The shelves in the store are stocked
according to a web of information systems.  The vast commodity market,
the "magic hand" that stabilizes this process is but a brittle skein of
interrelated information systems.

How many of these systems will be whittled by Y2K termites -One
percent? Five percent? More?  Which of the weakened links are critical
in holding society's meshwork of value chains together?


People who have been studying the Year 2000 problem are concerned
about the system that brings food to our tables, the payment systems that
support this process, and the financial systems that underlay stock and
bond markets.  They are also concerned about the systems that bring us
gasoline for our cars and run the electric power grid, the systems that
heat and cool our buildings, our healthcare system from HMOs to heart
monitors, and the systems that deliver government functions like Social
Security, Air Traffic Control, and local police, fire and ambulance

In each of these systems, information cascades from one system to
another, and another.  The date is often critical, though sometimes in
less-than-obvious ways.  For example, the date might be used as the seed
for a random number algorithm.  Or a date field like 9/9/99 might be
used to indicate the end of a list of numbers.  Each component might
depend on the correct operation of the systems that logically precede it.
Failure of one program has the potential to disrupt the superordinate
system to which it belongs.


Communications is a critical concern.  Microprocessors run virtually
every telecom system.  Arthur Gross, the former CIO of the US Internal
Revenue Service, is less concerned about the hundred million lines of
IRS application code than he is about the IRS's communications
capabilities.  "The Achilles heel for IRS . . . is going to be its
telecommunications problems," he said (4/14/98).

The Gartner Group's Howard Klein states publicly that customer premises
equipment is a critical concern.  He projects that at year-end 1988, 80%
of private telephone switches (PBXs), automatic call distributors
(ACDs), and interactive voice response (IVR) systems will not be Y2K
ready (10/1/97).

Bell Atlantic is less quantitative, but more dramatic: "Network
management tools and network hardware, such as bridges and routers,
built before 1996 may have been programmed with two-digit date fields
that will not respond to the Year 2000 change, causing networks to crash
at the dawn of the new millennium. If the Year 2000 problem is not
addressed from a network perspective, network administrators will be
unable to access mission critical applications. It is estimated that
networks installed before 1996 have a 90 percent chance of having a
Year 2000 problem" (Bell Atlantic Network Integration News Release,
November 3, 1997).


Yet, when it comes to the public telephone network, Bell Atlantic
assures us, "We're working systematically to make sure that January 1,
2000 is just another day for the (public) network" (http://www.bell- And AT&T says, "We don't
anticipate any Year-2000 related problems in providing service to our
customers before or beyond January 1, 2000"

Let's take it at face value, then, that the public switched network will
remain intact and functional.  But for heuristic value, let us assume that
the Year 2000 problem will hose a significant proportion of customer
premises systems.  And let's also assume that for some period(s) of time
beginning in 1999, and continuing into the Year 2000, that the operation
of the airlines, the Social Security System, the banking system, etc., is
sporadic and uncertain.

Now imagine yourself wondering whether your scheduled airliner will
fly tomorrow.  What would you do?  Myself, I'd call the airline to
find out.  And if I could not get through, I'd try again.  Hey, this
is important!  I'd keep trying until I got an answer!

What if your paycheck were delayed?  What if your mortgage company
screwed up your mortgage account? What if your local supermarket's
cash register treated your credit card as if it had expired? What if
the truck that was supposed to deliver your heating oil had not arrived?
What if you wanted to find out which day of the week your bank would
reopen?  Or whether the store had bread?  Myself, I'd call.  And keep
calling until I got through.

Yet, according to the Gartner Group, up to 90% of the private switches
and automatic call distributor systems could be down. The very
equipment that airlines and banks and mortgage companies and
government agencies use to handle customer calls might not be working.
For sure, they would try to handle incoming calls manually, as best they
could.  But we could expect a marked decrease in the number of calls
that could be handled. Even as the number of calls would be skyrocketing.


We're assuming that the public phone network is fully functional.  But
the phone network is engineered based on scarcity. We have already seen
single events (like the beginning of ticket sales for a Rolling Stones tour)
cause major network congestion. The fact that the phone network is
working does not mean that it would be problem free!

Your local telephone office is engineered for maybe one line in ten to be
busy during the Busy Hour.  What if this assumption is violated?  Maybe
a single bank emergency or spot shortage would generate assumption-
busting call volumes and maybe not.  Two such events, with the same
distal cause (Y2K), occurring at the same time in the same
neighborhood, would be more likely to.  Maybe you try to call and
can't get dial tone.  Or maybe you call and get network-busy from
the other end.  In either case, the network would work, but calls
might not go through.

So here's one way that "external" events might cause a working network
to become non-functional. Its an emergent problem.  There are many other
such emergent problems that could occur when a single trivial, but
pervasive factor, like the Year 2000 Problem, leans hard on the entire


Reports of tests, in which the clock is experimentally advanced to the
last minutes of 1999, are beginning to emerge.  An off-line power plant
failed at one minute past experimental midnight - a sensor that
integrated over time suddenly detected an out-of-range condition as the
denominator went from 99-something to 00-something.  A car factory
near Detroit that set its clocks ahead not only lost its time clocks,
but its security system failed - nobody could get in or out of the plant.

Another type of real-world evidence comes from a venerable, established,
high-tech corporation that spent $100 million and three years on a
software that was to make the company Y2K compliant -- and then abandoned
it in late 1997, because the fix was not matched to the corporate culture.

There are not very many such stories.  One reason is that the line
managers responsible for Year 2000 fixes come from a system where a
key goal is to make "your people" look good to your boss, where
showing positive results is more important than emphasizing the
challenge ahead.  Another is that corporate lawyers are advising their
clients to remain silent about the problem.  A third is that there simply
has not been very much "advance-the-clock" testing.


It is very unlikely that we will know which systems will fail, or what the
second-order consequences of those failures will be, or how people will
react, until the failures begin to occur.  When?  They are beginning to
occur - slowly - now.  We can expect a big pulse of failures around
January 1, 1999, and additional emergencies throughout the next year
or two.

We can surmise that telecommunications systems based on Stupid
Network principles will fail less frequently, and less severely,
than those based on Telecom Classic.  There are perhaps three
reasons for this.  First, these systems are newer, thus less likely to be
developed in the era of scarce memory.  Second, they are based on fewer
assumptions, and are easier to design by over provisioning.  Third, the
Internet Protocol was designed to be so that pieces of the network could
work even as other pieces fail.

But there is no call for complacency.  The global Internet is tied
at many points to older, more failure prone systems.

Our thought experiment, above, assumes that the public switched
telephone network survives with full functionality.
But there is reason to believe the opposite.
Extrapolating Capers Jones' data on on-time
software projects to AT&T's 500,000,000 lines gives us a minus 20-
something percent chance of project completion.  Whatever that means.

The five RBOCs and three other major US interexchange carriers have
software problems like AT&T's.  Then there's integration testing - all
telephone companies interconnect with each other, and their Y2K fixes
must interoperate.  AT&T interconnects with at least 280 other phone
companies worldwide.

As the critical date nears, we'll need to expect the unexpected.  And plan
for the unexpected, insofar as possible.  FCC Chairman Kennard recently
reported to Congress (April 28, 1998) that Y2K, "has the potential of
disrupting communications services worldwide."  Here's hoping that such
disruptions, if they occur, are of limited duration and severity!


On the other side of the coming crisis, perhaps (given good preparation,
enlightened leadership, and an otherwise peaceful world) we will be able
to look back and say, "We survived, we learned, we never abandoned our
humanity no matter how bad it got.  And our software infrastructure -
indeed our society - is better for it!"

(The best thinking on emergent Year 2000 issues and actions
that I'm aware of is on Doug Carmichael's website, . . .)

The Dawn of the Stupid Network was the cover story of the Feb/Mar 98
issue of ACM Networker (just out).  It is a major post-AT&T rewrite of
the Stupid Network idea.  It'll be on soon.  Better
yet, AT&T Executive Vice President and all-around good guy Dado Vraslovic,
is writing a rebuttal, defending intelligence in the network, for the
April/May issue of ACM Networker, due out RSN.

SMART REMARK OF THE MONTH:  "If I taught a class, on my final exam
I would take an Internet company and ask [my students], 'How much is
this company worth?' Anyone who would answer, I would flunk."
Warren Buffett, quoted in the Industry Standard Media Grok, May 5, 1998.


On May 18, I will be lead-off speaker for a Washington DC
conference called "The Bandwidth Explosion: Understanding New
Technologies That Are Driving Business Opportunities." The website

And on May 19, also in Washington DC, I will address the OIDA
Communications Roadmap Workshop.  OIDA stands for Optoelectronics
Industry Development Association.  The title of my talk will be
"Stupid Optical Networks: A Bright Idea."
Find OIDA contact information on

On June 8-11, I will be at Jeff Pulver's Voice-on-the-Net Europe
Conference.  I don't know what I will talk about.

On the evening of June 11, in London, I will speak at a World
Technology Network roundtable.  A cozy dinner discussion with
technology leaders from the UK.  SMART people are welcome.
Contact for more info.

On Thursday, July 9, I have been engaged as Agent Provocateur at
a Washington DC Telestrategies conference entitled "Next Generation
IP Networks."

DON'T MISS George Gilder's TELECOSM Meeting, September 15-17, Lake
Tahoe NV.  Last year AT&T would not let me speak at Telecosm.  This
year, not only will I be speaking, I'll be hosting a panel of
other AT&T refugees!  Tom Evslin and Joe Nacchio have accepted slots
on my panel, so far.  Watch this space for additional surprise guests
(as they leave AT&T).  Info at . . .

David I
<<to unsubscribe to the SMART List, send a brief unsubscribe message

<<for past SMART Letters, see>>
David S. Isenberg     
18 South Wickom Drive   888-isen-com (anytime)
Westfield NJ 07090 USA  908-875-0772 (direct line)
                                908-654-0772 (home)
     -- Technology Analysis and Strategy --
        Rethinking the value of networks
      in an era of abundant infrastructure.

Date last modified: 13 May 98