The Year 10,000 problem (also known as the Y10K problem or the deca-millennium bug[1]) is the class of all potential software bugs that would emerge when the need to express years with five digits arises. The problem can have discernible effects today, but is also sometimes mentioned for humorous effect as in RFC 2550.
Contents |
Historical and technological trends suggest that in the actual year 10,000 it is extremely unlikely that any of the data processing technology or software in use today will still be active, nor can it be known whether the present Gregorian calendar system will even still be in use. However, five-digit years are already a problem today for some forward-looking analysis programs, such as software that examines proposals for the long-term handling of nuclear waste.
Many date-handling programs or routines created prior to the recognition of the Y2K problem only used the last two digits of a year for storage and calculation, such as "60" for 1960. The routines would then either add the constant 1900 to the result, or even just insert the text string "19" in front of the result when displaying it, causing the year 2000 to be displayed or interpreted as 1800, 1900, 19100, or 100. Although this was often done to preserve what was at the time precious storage and memory space (reasons now unlikely to be relevant), it was also done as a mirror of cultural practice (saying "the '60s" instead of "the 1960s", for example) or to convey information in a limited display space.
This problem can be seen in the spreadsheet program Microsoft Excel through at least the Office Excel 2007 release, which stores dates as the number of days since 31 December 1899 (day 1 is 1900-01-01), and the database program Microsoft Access, which stores dates as the number of days since 30 December 1899 (day 1 is 1899-12-31). In either application, a date value of 2958465 will be correctly formatted as "31 December 9999" but adding 1 to that to step over to the expected date of "1 January 10000" will cause a formatting error; in Excel 2000, for example, it will be displayed in the cell as a series of # characters. Excel also cannot automatically convert date-formatted strings such as "12/12/2007" to dates if the year exceeds 9999; "12/12/9999" is automatically converted to a date when entered into a cell, but "12/12/10000" is not.
The Long Now Foundation ran into this limitation of Excel during the design of the 10,000 year clock.[2]
The open source OpenOffice.org Calc program is able to display dates beyond the year 9999 correctly with five digit years, but at least through version 2.4 falls victim to the Year 32,768 problem: "31 December 32,767" is the highest available date it can properly display. 32767, or 215-1, is the highest positive number that can be represented using a 16-bit signed integer, adding one to this value causes it to overflow, and Calc interprets the year as a large negative number, "1 January -32768".
The GNU Fortran compiler, g77, makes reference in run-time environment limits to year 10000 (Y10K) problems when using intrinsic functions with this compiler suite. The problem is simply stated as, "Most intrinsics returning, or computing values based on, date information are prone to Year-10000 (Y10K) problems, due to supporting only 4 digits for the year." The failure mode suggested in all of the intrinsic functions is that, "Programs making use of this intrinsic might not be Year 10000 (Y10K) compliant. For example, the date might appear, to such programs, to wrap around (change from a larger value to a smaller one) as of the Year 10000."[3]
Unlike the Y2K problem, where significant digits were omitted from the stored values of years, fixing the Year 10,000 problem does not require updating old records (assuming they are already Y2K compliant), since all four significant digits are present. It only requires that record storage in decimal be able to store five or more digits.
There is, however, a potential problem with record sets that make use of lexical sorting. For example, representations of dates in the range 10,000-19,999 would appear interlaced with dates in the range 1000-1999 rather than after the year 9999.
The Long Now Foundation is attempting to foster the custom of writing years with five digits, so that the year 2000 would be written as "02000". This would preempt the Year 10,000 problem, but would in turn be susceptible to a "Year 100,000 problem".
The Internet Kermit Service Daemon (IKSD) uses a five-digit field for the year in the Database Record Format: "Date-time fields are right-adjusted within a field of 18 with the leading blank reserved for Y10K".[4]
ISO 8601 specifies that years be written with four digits, but allows for extension to five or more digits, with prior agreement between the parties exchanging the information.
|
|||||||||||