RISKVUE ARCHIVE | RISK BITES

Hot Dates And Your Computer

By Norman Ritchie

Y2K is only the beginning. There are a large number of dates which may cause some degree of difficulty to microprocessor based systems. These dates are similar to the well-known Y2K Problem, but differ significantly in scope and potential impact. Some of these dates have already passed with scarcely a murmur heard outside of the IT community. The following list has been compiled by Risk Solutions from a large number of sources, including their own consultants and many contributions obtained from the Internet.

Definitions

A few definitions will help the less-nerdy reader understand this list better. Complete understanding of all of the dates will be difficult unless you are deeply involved with computers and computer programming. There is some extremely arcane programming artistry behind the whole issue of problem dates.

Signal Entry refers to the programming practice of using a particular number as an indication to the software that, for example, it has reached the end of a file or that the entry is not real data.

Nonsense Dates are necessary entries in some software because all date fields must have some content. When a “nonsense” date is entered, the software normally recognizes it as such. A problem may occur when the nonsense date is a real date—the software may not take it seriously!

Field Rollover occurs when the bit length of the binary number form of a date exceeds the maximum number of digits allowed by the computer. The field then rolls over to zero causing problems to software. This is also known as field overflow.

Even now, there is not a full understanding of all the dates that may affect computer hardware and/or software. Some of the dates in this list are still controversial, even amongst programmers.

Dates to Watch — 1999 to 10000 AD

• 1998/07/06 — Possibly “9876” might be a “signal” entry
• 1998/09/09 — Sometimes a “signal” date — cf. 1999/09/09
• 1999/01/01 — Start of the last year of the 1900s; “99” may be a “signal”
• 1999/01/01 — Effective introduction of the Euro (ex-ecu) within contiguous Europe. The timing of the Euro’s introduction is problematic because it drains resources from Year 2000 activities.
• 1999/04/01 — Start of Canadian & Japanese FY including 2000
• 1999/04/06 — Start of United Kingdom FY 1999-2000
• 1999/04/20 — 255 days to year 2000
• 1999/07/01 — USA: 46 States’ fiscal Y2K starts, FY 2000. But NY=04/01, TX=09/01; AL, MI=10/01.
• 1999/08/22 — End of GPS (Geographical Positioning Systems) Week 1023 @0000h, 10-bit field rollover
• 1999/09/09 — Default “nonsense” date in many data-entry screens. Also, number string “9999” has been used as end of file indicator.
• 1999/09/23 — 99 days to year 2000
• 1999/10/01 — USA’s federal government fiscal Y2K starts, FY 2000
• 1999/10/03 — 90 days to Year 2000 — 90 day intervals are common (& 60,30)
• 1999/12/** — A period of maximum solar activity which may affect communication devices and systems (may need to be considered in relation to contingency planning)
• 1999/12/01 — Dec ’99 may be a “Signal”
• 1999/12/31 — Sometimes used as “Never Expires” date (IBM Tapes marked 99365—all may expire today, or not; also 99366, 99999). Support for much software may cease after today.
• 1999/99/99 &c — a better “nonsense” or “special” date than Sept 9th. May be an “end of file” indicator
• 2000/##/## — Still 20th Century, but some will say 21st
• 2???/??/?? — Probable introduction of the Euro in the UK. The timing of the Euro’s introduction is problematic because it drains resources from Year 2000 activities.
• 2000/01/01 — Date-form is longer than ever before—field overflow is possible
• 2000/01/03 — Monday—First Business Day of 2000 (USA). Important for contingency planning.
• 2000/01/04 — Tuesday—First Business Day of 2000 (UK). Important for contingency planning.
• 2000/01/05 — Date-form is longer than ever before—field overflow is possible
• 2000/01/10 — Date-form is longer than ever before—field overflow is possible
• 2000/01/12 — Date-form is longer than ever before—field overflow is possible
• 2000/02/29 — Tuesday—Not everyone is aware that 2000 is a leap year and this valid date is not expected to occur by all software. The date may not be accepted by software or may cause errors or application failures.
• 2000/03/01 — Some leap year errors may not have shown yesterday
• 2000/09/01 — Date-form is longer than ever before—field overflow is possible
• 2000/09/06 — Date-form is longer than ever before—field overflow is possible
• 2000/09/09 — Sometimes a “signal” date as per 1999/09/09
• 2000/09/10 — Date-form is longer than ever before—field overflow is possible
• 2000/09/13 — Date-form is longer than ever before—field overflow is possible
• 2000/09/27 — Date-form is longer than ever before—field overflow is possible
• 2000/10/01 — Date-form is longer than ever before—field overflow is possible
• 2000/10/04 — Date-form is longer than ever before—field overflow is possible
• 2000/10/10 — Date-form is longer than ever before—field overflow is possible
• 2000/10/11 — Date-form is longer than ever before—field overflow is possible
• 2000/12/31 — 366th day of year. May not be expected by systems which did not acknowledge the leap year.
• 2001/01/01 — Third Millennium and Twenty-First Century start
• 2001/02/29 — Will not exist; but check software
• 2001/09/08 — In some UNIX applications 999,999,999 has been used to indicate an end of file situation. In UNIX, the number string of 999,999,999 is actually the date of September 8, 2001.
• 2003/01/01 — Burroughs Unisys A Series system rollover date
• 2004/02/29 — First “Leap Day” of c21 & M3
• 2004/08/01 — UK car number plate year letter would run out, on present scheme
• 2005/??/?? — Some really old versions of UNIX (e.g. 16-bit BSD) die in 2005
• 2012/12/2? — See 2023/12/23
• 2020/01/01 — Henceforth, Mac (System 6.0.4+) Date & Time control panel cannot set current date
• 2023/12/23 — End of World, Mayan Calendar (Sunday). Maybe 21 or 23 December AD 2012
• 2024 — May contain a date when there will be an overflow problem similar to 2038 below, but the precise date and the cause have not been ascertained
• Circa 2025 — In the US, the number of 10 digit telephone numbers begins to exceed the overall capacity of the number of digits available (projected)
• 2030 — This has been reported as a breakpoint in the windowing system used by Microsoft in a large number of their products. In these systems 29 will imply 2029 and 30 will imply 1930.
• 2036/01/01 — Burroughs Unisys A Series system date
• 2036/02/06 — NTP timestamp overflow (2E32 secs from 1900/01/01-00:00:00 GMT)
• 2038/01/19 — End of UNIX time: (Tue) 2E31 secs from 1970/01/01, time MSB sets @03:14:08 GMT. The UNIX clock reaches 2,147,483,647 which is the largest number which can be stored as 32-bit signed integer and will roll-over at 3:14:07 (at which point it will seem to be 1970 again or at least the number of seconds from January 1, 1970 will seem to be 0).
• 2038/04/23 — MJD 2E16
• 2040/02/06 — At 06:28:16, old Macs’ longword seconds from 1904/01/01 overflows
• 2042/09/17 — IBM 370 TOD clock overflow (long seconds (1.048576 s) from 1900)
• 2044/01/01 — MSDOS : 2E6 years from 1980, file date MSB sets
• 2046/01/01 — Amiga system date failure
• 2048/01/01 — AD 2E11 starts—12-bit signed overflow—PDP-8???
• 2048/01/19 — Stratus VOS OS failure (1980/01/01 2E31 seconds)
• 2048/06/?? — Some UNIX password ageing fails; 64E2 weeks from 1970
• 2049/12/31 — Microsoft Project 95 limit
• Circa 2050 — The capacity of the social security number system of about 1 billion unique numbers is exceeded extrapolated from current usage rates
• 2078/12/31 — Excel 7.0 & Excel 95—The Last Day, #65380
• 2079/06/06 — 2E16 days from 1900/01/01
• 2080/01/01 — MSDOS file dates, when displayed with two-digit years, are ambiguous if files may be dated today or later. Windows File Manager, set to ISO-8601 dates, drops 100 years from displayed file dates of 2080.
• 2100/01/01 — Y2.1K. Most current PC BIOS run out of dates. MSDOS DIR renders filedate years 2100-2107 as 99
• 2100/02/29 — Next non-existent February 29 in a year divisible by 4 (except for Greek Orthodox outside Greece, using Julian); End of 200 years of 28-year calendar repeat. Start of 100 years thereof.
• 2101/03/01 — It is claimed that a system which is compliant up to this date will be totally and permanently date compliant
• 2106/02/07 — UNIX : 2**32 secs from 1970/01/01, time overflows @06:28:16
• 2108/01/01 — MSDOS : 2**7 years from 1980, file date overflows
• 2132/07/22 or
  2132/09/01 — JD 2500000 starts at noon; 12 hours to MJD sixth digit
• 2135/01/01 — MMDCCCLXXXVIII : the Roman A.U.C. date field will be longer than ever before
• 2247/01/01 — Y3K A.U.C. starts
• 2400/02/29 — will exist—The 100-year rule is overridden
• 2800/02/29 — first difference between Gregorian and Greek Orthodox
• 2888/01/01 — MMDCCCLXXXVIII : the Roman A.D. date field will be longer than ever before
• 3268/01/01 — (Julian date—3268/01/24 Gregorian?) start of Second Julian Period, noon UTC
• 4000/01/01 — Roman year number is ill-understood after MMMCMXCIX.
• 4000/02/29 — Allegedly, 4000 is not Leap in the former USSR.
• 4338/11/28 — Cobol-85 Integer day 1000000—exceeds 6-digit field.
• 9999/01/01 — HTTP caching fails (or at year’s end?).
• 10000/01/01 — Y10K

Repeated Risks

• 2000/01/(01-M) — M days to 2000, for any special short-term M (30 days, ...)
• ####/01/01 — N years to/from 2000—possible problems with N-year subscriptions
• ####/##/## — N years to/from your FY 2000—fiscal year 2000
• 20#0/01/01 — Likely date windowing critical points
• Clock base dates — UNIX 1970/01/01, PC 1980/01/01, etc.
• ANY date with a memorable pattern late in the 1900’s may have been chosen in the past as a “signal” date
• Autumn — clocks go back in EU & NA—one hour’s worth of civil time is repeated, timed events may be duplicated. VCRs, ATMs, etc may not be programmed to recognize this. Windows 95 sets the clock back whenever it thinks that the clock has reached 2am on the day in question, not just on the first occasion.
• Spring — clocks go forward in EU & NA—one hour’s worth of civil time is omitted, timed events may be skipped 

Source: Risk Solutions, a company specializing in the quantification, control and management of industrial risk. Risk Solutions have considerable experience in providing services and advice to a number of industries, and a track record of increasing safety, availability and profitability for our clients.

Risk Solutions applies its expertise to all stages of an investment life cycle from conceptual design, through existing plant and operations, to abandonment, removal and remediation. With particular regard to insurers and investors, the company’s expertise is a valuable tool in setting accurate levels of coverage by identifying unsafe or inadequate design, procedures, operations or maintenance and ensuring that either risk reduction measures are undertaken or that insurance is loaded to reflect the actual risk exposure.

riskVue | The webzine for risk management professionals
September 1999