What's so hard about the Year 2000 problem? Senator
Bob Bennett compared it with changing all the rivets on
the Golden Gate bridge. Changing one rivet is easy,
but changing all of them while keeping traffic moving
is a tremendous challenge. Another analogy is repairing
a Boeing 747 while in flight. Y2K has also been compared with
replacing light bulbs in Las Vegas in one afternoon,
or to requiring a visual inspection of every vehicle part
in an entire town. Or maybe you relate to resoldering
all of the plumbing in the US, but with a limited supply
of plumbers. (That pretty well conveys the costs involved.)
More accurate -- but boring -- is an analogy with reviewing
every book in every library to convert dates to a new format,
and trying to coordinate the changes across all libraries
so that no errors slip through. [Rod Swab The Y2K problem is complex because there are so many
different date formats and ways of doing date arithmetic.
Some computers will wrap 99 to 00, others will increment it
to 100, and still others will signal an error or fail in
non-intuitive ways. Some work with a 2-digit date directly,
and may be unable to handle 00 or 100, others will concatenate
'19' to get 1900 or 19100, others will add 1900 to get 1900
or 2000. Global Positioning System (GPS) week numbers
will wrap from week 1023 to week 0 in Aug99. On many PCs,
the clock will go from 31Dec99 to 04Jan80, and on some of them
there is no way to force the PC to recognize Year 2000 dates.
As many as half of all PCs will have occasional RTC/BIOS
communication problems -- especially with slower processors
-- and even some Macs suffer from this Crouch-Echlin Effect.
(See Some systems know that 2000 is a leap year, but many do not.
Any errors will be propagated through sorts, comparisons,
and calculations, resulting in unexpected actions, invalid data,
and garbled reports. Testing for these problems is tricky
because setting your computer clock ahead can trigger license
expirations, expire your passwords, or corrupt your databases,
and may even make your PC unbootable. See
Incidentally, some Macintosh computers can't handle
dates after 2019 and Unix has an overflow problem in 2038.
For a list of critical dates, see
Of course, for each system that fails there will be people
hard at work to bypass the problems and restore basic functioning.
People with phasers cutting through bulkheads or crawling through
Jeffries tubes. But keep in mind that computing systems
are ubiquitous and interconnected, and that many of these people
are the same ones that help desks have to tell repeatedly
to check the power plug or turn on the computer instead of
hitting it with a shoe. (Bill Park recommends Dave Pogue's
little paperback recounting real computer help-line calls.)
Even if they get something fixed, they have to make sure
it stays fixed and that the fix evolves to accommodate
changing fixes to other related systems. And all of this is
very hard to do if parts are unavailable or the power
and communications grids are down -- even if what you're
trying to fix is a part of the power grid.
For a long, detailed, and depressing report on what
might happen at the turn of the century, see the Sunday NY Post.
"Experts warn it could wreak havoc on Jan. 1, 2000 -- triggering
a worldwide paralysis." Corporate analysts are saying that their earlier Y2K estimates
(totalling $600B worldwide) are too low. Y2K was seen as
a COBOL/mainframe problem, but it's evident that PCs, networks,
and distributed systems will take a nearly equal effort.
(The few desktop tools available suffer from "release 1.0
syndrome," and even an upgrade to Windows NT can take 45 minutes
of on-site technical work per PC. Authority for desktop systems
is often decentralized, and there's no one responsible for
checking all the spreadsheet macros.) Gartner Group now estimates
average Y2K costs of $6.46 per line of code -- nearly six times
their original estimate of $1.10. That's allowing for
all languages, PCs, servers, embedded systems, networks,
project management, supply-chain analysis, etc. There may be
50M embedded devices with Y2K problems, and only 5% of large US
companies have started worrying about it. Embedded systems
are often inaccessible, have little documentation, have few
testing and development tools, and use code that was not
written in-house. (The compilers may no longer even exist.)
Vendor assurances of Y2K compliance can't be trusted --
if they even bother to respond -- and in many cases
the vendors are no longer in business.
The US government is facing a shortage of programmers
willing to work on Y2K problems at government salaries.
It is estimated that more than 40% of critical systems will not
be fixed by Mar99. Any breakdown of FAA computers could
affect air traffic, including mail and freight delivery.
Delays in Medicare payments would affect hospitals.
Malfunctions in security systems could lock workers out of
their computers and labs. [Rajiv Chandrasekaran, Washington Post.
SJM, 03Jun98, 1C.]
Private sector recruiters are in a feeding frenzy.
Y2K programmers are leaving the government for high-paying jobs
where they won't be expected to solve problems with ancient
systems and embedded chips. Embedded systems are particularly
difficult because you can't duplicate the system environment
for testing, and in many cases can't roll the date forward
without destroying the equipment. Large telecom operations
with thousands of hubs and routers cannot be tested simultaneously
as they operate in the real world. CalTrans has found that
2% of its chips are going to be a problem, particularly those
in buildings put up or remodeled between 1985 and 1993.
Access to the chips is difficult, vendors are often unreachable,
and testing can take four hours per device. Most companies
have yet to develop the specific test procedures they need.
Agencies in charge of critical systems are talking about
"contingency planning" and "disaster recovery," but few have
comprehensive plans. HP3000 COBOL programmers in Silicon Valley bill at
about $80-$100/hour, with project managers and consultants
getting more. (25%-40% of the billable rate may go to an agency.)
The rate has gone up $10/hour from last year, and $20/hour
from a decade ago. COBOL programmers seem to be in their
late 30s and above. There are rumors of pay rates elsewhere
of $120/hour, perhaps going to $150/hour. [Chris A. Goodey
The new Y2K cost estimates will affect effort and resources
required, time to completion, and of course budgets. Citicorp's
estimated $600M cost, for instance, would now be more like $3.5B.
Federal estimates of $5B may escalate to $30B-$50B, of which
they've only spent $1.5B so far. When the crisis comes,
we're going to lose power, water, transportation, communications,
and other services. (Also manufacturing, which depends heavily
on embedded systems.) "The wise will get out of populated areas
while they still have a chance. The Federal Government will not
feed you. The State will not feed you. The County will not
feed you. Your Town will not feed you." [Paul Milne
It may be wise to stock up on candles, food staples,
potable water, hard currency, fuel, and batteries, or even
move to a farm. Obtaining paper copies of important records
is also a good idea. [Pam Hystad I asked analyst Mark Anderson where one might invest
retirement funds in order to avoid a coming Y2K stock market
correction. He noted that people will always need food, shelter,
and medicine, and recommended food as a very conservative sector.
[ Incidentally, the lawyers are doing their part to
save the world economy. Intuit is now facing a Y2K lawsuit.
-- Ken