Greenwich Mean Time

August 1998

In May 1997 Greenwich Mean Time released the Hindhead Report which detailed the high percentage of year 2000 date related problems in the world’s best selling PC software programs (64.2%) and hardware systems (93%). (Appendix 1.) As our independent research has continued, we have added many thousands of software programs and hardware systems to our knowledge base and deepened our understanding of the year 2000 PC problem.

Despite the wealth of collective year 2000 wisdom in the public domain – and the tools to assist with problem resolution – large corporates and government departments have been slow to act on year 2000 problems. Small and medium enterprises have been even slower. Yet there are 5.6 million businesses in the USA that employ 500 or fewer people. (Source: Bureau of the Census).

Like the USA, most countries rely on a backbone of small and medium enterprises. Their continued inaction poses a substantial threat to world economies.

As August 19 approaches - 500 days until 2000 - we present the Alpine Report. It is based on the Hindhead Report, our ongoing research, and the practical experiences of thousands of businesses over the past 2½ years.

This case study illustrates the fact that there is still time to resolve the problem. And while the costs we have detailed for a year 2000 PC project may seem high for small and medium sized enterprises, the alternative is the highest cost of all: business failure within the first quarter of 2000.

A typical year 2000 PC project in a 500 person company:

• Requires 467 man days to overcome problems;
• Costs in the region of $845,800

The year 2000 problem occurs within all five layers of the PC: hardware, operating systems, software programs, user data, and data sharing. (See Appendix 2.)

• Less than 1% of the cost is spent on hardware year 2000 issues.
• Just over 1% of the cost is spent on operating systems issues.
• 21% of the cost is spent on software issues.
• 77% of the cost is spent on data issues (including data sharing issues).

A soft drinks manufacturing and bottling company which employs 500 people.

• Breakdown of staff by department and PCs per department is representative.
• All PCs were replaced in first half of 1997, reflecting a 3 year replacement cycle.
• Software used is typical common off the shelf software (COTS) packages.
• Each PC has on average 50 spreadsheets or databases, of which ¼ have some year 2000 issues.
• All office staff use a PC network.
• Phases of year 2000 risk are diagnose, fix, reconfigure, replace or upgrade, and retrain.
• 75% of operating system and software program issues can be resolved with reconfiguration, the remaining 25% will require upgrading.
• Cost of hardware fix will be $10 per PC and will take 5 minutes.
• Cost of operating system fix will be $15 per PC and will take 15 minutes.
• Cost of reconfiguration of software programs will be $25 per PC and will take 15 minutes per problem package.
• Cost of replacement of COTS software will be $100 per problem software package and will take one hour per package to install and configure.
• Cost of fixing non-compliant data will be $100 per problem data file.
• Retraining costs are $100 per hour (excludes opportunity cost and lost production).
• Retraining users on new software packages will average 4 hours per new package.

29 May 1997

Revised 22 October 1997, 31 July 1998

SOFTWARE PROGRAMS

Of 4000 common off the shelf (COTS) software programs checked by Greenwich Mean Time, 2568 (64.2%) exhibit some sort of year 2000 date dependent behaviour. Of these, 724 (28.2%) are claimed to be year 2000 compliant by their manufacturers.

It is important to understand how this anomaly arises.

When software manufacturers claim their products are year 2000 compliant, they generally mean that they may somehow be configured to have no date dependent behaviour.

During installation most software programs give the user a choice of either standard or advanced installation. Advanced installation is usually recommended for experienced users only, so most people naturally choose standard installation. However, to make the software operate in a compliant manner, the user often needs the advanced installation. Most users don’t know this and, even if they did, they‘re unlikely to know how to use the advanced options.

Greenwich Mean Time remains the only company in the world today to have defined year 2000 issues. Among the 2568 programs tested for the original Hindhead Report, these are the problem definitions:

• 24.5% (629) manipulate dates to perform date calculations. This can create problems where the dates being used are in a different century from the one which the program infers.

• 21% (539) are programming development tools which are capable of creating programs that use dates incorrectly, so perpetuating non-compliance problems.

• 19.7% (506) use a process known as "date windowing". While at first glance such programs seem to be year 2000 compliant, they create indirect problems leading up to 2000, as the program must decide which dates belong in which century. Which century the program chooses is often not apparent. Furthermore, these date "windows" differ from one program to another – and even from version to version of the same program - and the user has no way of changing them.

• 11.5% (295) store dates differently from the way the user inputs them. These programs can work in 2000 but the user has no control over any function to do with date computations.

• 9.9% (254) have licensing issues which means that there is a time bomb embedded within them to counteract illicit use of the software. If the software manufacturer did not cater for either the year 2000 or the year 1900, this time bomb may go off unexpectedly.

• 5.5% (142) import dates from either the operating system or the BIOS, either of which may not register 2000 correctly.

• 4.4% (113) do not recognize 2000 as a leap year (which it is), creating day-of-week errors from 29 February 2000 onwards.

• 3.5% (90) of COTS programs that are commonly used today will only function in the 20^th century. Users need to know which these are so they can abandon them in favour of programs that do work in 2000.

These figures represent a best case scenario. Most of the programs tested exhibited more than one category of year 2000 issue. The total number of issues found among the 2568 programs was 7148. In defining the issues, Greenwich Mean Time selected only the one issue per program that would have the biggest impact on users of that software, and counted this as the primary year 2000 issue.

HARDWARE

At the hardware level the BIOS (basic input/output system) has been the issue as it initialises the PC, including setting up the system date and time by reading values from the real time clock chip which keeps a constant time reference for the PC. If the BIOS does not successfully rollover to 2000, all date and time references will be incorrect.

Greenwich Mean Time found that 93% of 1996 and pre 1996 BIOSs do not rollover successfully to 2000. Of these, 79% fail by rolling to 1900 after 1999, and 14% fail by rolling to some other date.

This figure drops substantially for BIOSs created after 1996 to 47% which do not rollover successfully to 2000, and to 21% by the end of 1997 and 11% for the first half of 1998. However, 98% of faulty BIOS types can be easily corrected.

The danger lies in the high proportion of companies that believe the BIOS is the full extent of the problem. While it must be addressed because it passes date and time information to operating systems and some software programs, Greenwich Mean Time defines it as a mere 1% of the overall year 2000 PC problem.

The BIOS initialises every PC on start up and passes date and time information on to operating systems and some software programs. Some BIOS types don’t rollover correctly as a result of a two digit year from the real time clock chip being incorrectly expanded to a four digit year by the BIOS. Approximately 98% of BIOS types can be corrected.

The operating system is sometimes responsible for feeding information to the other layers. The standard installation of most common PC operating systems is not optimised for 2000 but, using advanced installation, most operating systems can be configured to deal with 2000. In most common network operating system environments some reconfiguration will be necessary and this may involve replacement on every PC.

A high percentage of PC software programs exhibit potential year 2000 problems in normal usage mode. Greenwich Mean Time has so far identified 73 different categories of problem and severity levels, the most serious being "date windowing" which allows the PC to guess the century when only the last two digits of the year are entered. Date windows vary from one software program to another and even from version to version of the same software program. They are generally not obvious to the user. Software programs either need to be corrected (where possible), reconfigured, upgraded, or replaced.

Data is fed into a PC where it is turned into information. Years are most often entered as two digits (YY), yet most PC programs need four digit years (CCYY) to perform calculations. Many automatically expand YY dates to CCYY dates, assuming centuries in the process. Different programs resident on one PC or in one company may assume different centuries and any number of these may be incorrect. This is not apparent to the user who makes decisions based on incorrect information, perpetuating the problem.

Faulty data needs to be located and corrected. All data must be assessed for the use of YY years and programs must be assessed for the acceptance of correct data handling. New data must be monitored to ensure that it is year 2000 compliant.

As data is exchanged - between programs, PCs, departments and companies - it is often modified by programs as they make their different century assumptions. Data is continually being exchanged by e-mail, the Internet, EDI, disk, and across networks. Even the common technique of cut and paste may result in year 2000 problems.

Data corruption is worse than computer failure as users are often unaware that data has been corrupted and continue to use it and base decisions on it. Problems that occur at all of the other layers are compounded when data is exchanged. Data sharing methods need to be corrected and monitored.