Sounding the net alarm on year 2000 problems

Call him a Chicken Little or a Paul Revere, but Karl Feilder is certainly shouting at the top of his lungs that the Year 2000 problem isn't only about mainframes.

PCs and client-server networks are at far greater risk than commonly believed, says Feilder, CEO of Greenwich Mean Time, Ltd., a British company geared to addressing that aspect of Y2K. Greenwich recently tested 4,000 popular software programs and found 64 percent to have a Y2K problem.

Feilder spoke last week with Network World Senior Writer Paul McNamara.

Q. Conventional wisdom holds that Year 2000 is a mainframe problem that network managers need not fret about. What do you think?

A. Year 2000 may not be of concern to the network managers, but it bloody well should be. At the moment they are trying to keep their heads down, but I have been to see organizations all around the world - between 10,000 and 160,000 PCs in some of them - and they are scared.

At one of the big banks in South Africa the Year 2000 manager has just resigned, because she has 56,000 PCs around the world and they will not make any decision as to what to do about them. She said, 'I don't want to be associated with this project or this company.'

If there is any organization that thinks there isn't a big Year 2000 PC problem, let them switch all their PCs off and see how their bloody company works.

Q. What do you envision as the real-life consequences if this problem goes unaddressed?

A. The most likely thing is that your e-mail system won't work, your payroll system won't work and most of your financial calculations will go wrong and they will go wrong to the extent that it will take you longer than you've got in the remaining life of your company to fix them. If you can't issue invoices and you can't pay your staff, how the hell do you think you're going to run your company.

Q. What might happen with e-mail?

A. If you don't fix your PC hardware, then (on Jan. 1, 2000) your PC is most likely to think that it's the year 1900. If I send you an e-mail dated 1900, it probably won't get to you, because somewhere along the line between me and you either a router or a mail gateway will look at that message and say, 'Hey, this message is a hundred years old,' and will just delete it. ... The more modern e-mail systems will cope, provided that the machines they're running on are compliant.

Q. What about receiving mail? Will that be affected?

A. It may be, yes. Because your mail software running on your PC would take its date understanding from your PC hardware, so if your PC hardware thought it was 1980, for example, it may deliver messages in the wrong order or corrupt it or just reject it. We've seen mail messages being corrupted, being miscalculated and missorted, based on the way that the PC is communicating with the mail server.

Q. What's the scope of the Year 2000 problem in PC hardware?

A. We have conducted tests on more than 500 BIOSs and we found that of the pre-1997 machines, we got a 93 percent failure rate on the Year 2000 and on the 1997 ones, we saw a 47 percent failure rate. It was a big shock to a lot of people.

Q. How could that be, given that the problem has been known for so long?

A. There is an assumption on the part of PC manufacturers that people will simply buy new PCs before the year 2000, which is not a realistic assumption, especially if you go into any corporation today that is still using machines that are four or five years old. And there is also an assumption on the part of the buyers at the corporate level that is a problem. They have not been putting any pressure on the manufacturers.

Q. What causes the PC hardware problem?

A. "The BIOS, in most cases, thinks that the year after 1999 is the year 1900. It passes 1900 to the operating system on the PC, which is your Windows or DOS or whatever you've got on the screen, and it's up to the operating system to interpret what it thinks that date means. DOS systems seem to think that that date is obviously a mistake and DOS systems report the 4th of January 1980; that's their error date. The consequence of that is any DOS-based system like your accounts package or payroll system will think it's the Jan. 4, 1980. Consequently, when you run your payroll, you try to do any billing, it will get the date wrong and screw up all your accounts.

On a Windows-based system, some of them just reflect the date as 1900, in which case the date is wrong and the day of the week is wrong, because the first of January 1900 was a Monday and the first of January 2000 is a Saturday. In those systems it just gets the date wrong and any programs running on the Windows system will think it's the first of January 1900. That may cause them to fail, because some of those programs are more intelligent and they say, 'It can't possible be the first of January 1900, I'm not going to run.'

Some of the more modern systems don't worry about whether the PC says it's 1900 or 2000, because all they look at is the year 00, and they say, if it's 00, it must mean 2000. On some of the more modern systems it's not that big a problem.

A lot has been made of the BIOS and hardware problem and really it's overstated (as opposed to the software problems). The problem occurs in a hell of a lot of machines, that is for sure, but fixing it is relatively straightforward. There are a number of fixes available on the open market and there are a number of fixes from the people who make the PCs. All you've got to do is know whether your machine passes or fails.

The worst thing about this is that you have to test every single PC, because every model of PC actually contains different components. We've seen cases where a brand new machine will fail and a machine that is one month old will pass.

Q. What happens at the O/S level?

A. The only really big problem there is that if your BIOS fails, then your operating system is going to report the wrong date in some cases.

Q. How do the applications fail?

A. Basically, the software very rarely absolutely fails and very rarely absolutely passes. We've categorized it into 30 different shades of gray between pass and fail. Somewhere in those 30 will be your definition of compliant and somewhere in that 30 will be the manufacturer's definition of compliant, and it's rare that the two are the same thing. With 30 different shades of gray, it's open to a lot of interpretation.

The biggest source of the problems when it comes to software is spreadsheets and databases. In most spreadsheets, when you enter a two-digit year, like 53, they make a pre-programmed guess as to what century you're talking about. The criteria for those guesses are hard-coded into the program, you can't change them. So if you take, for example, Microsoft Excel 97 and you type in the year 28, it assumes you mean 2028. If you type in the year 30, it assumes 1930. There is a magic date in Excel 97, which we call a pivot date, and the pivot date is 29.

There are a number -- our research shows 4 percent -- of current programs which will not run after Dec. 31, 1999, they simply just stop. Some of those programs are accounts packages, some are payroll systems.

The way around this is to use 4-digit years.

Q. So why would anybody do anything else?

A. If I gave you a job of typing in a hundred people's date of birth every day, you would very rapidly stop typing in the century.

The problem becomes: what do you do about all your old spreadsheets and databases? What do you do about all your old stuff that you still use today? Basically, you have to go through every one of those spreadsheets and databases that you still use and expand the two-digit year to a four-digit year. That's where people have missed the point, because this is going to take a bloody long time. There's no automated way of doing it.

Q. Care to name some applications that fail?

A. No thanks, not in the States.

Q. Why are vendors soft-pedaling the problem?

A. When you consider that most of the software companies are very rich and their motivation in life is based around their share price, you start to appreciate that even a hint that they might be open to some sort of litigation, even the suggestion that they might lose some of their customers, would dramatically effect the share price and hence they are avoiding making these sort of comments.

I am personally amazed that Wall Street hasn't noticed this yet, that the whole of the computer industry is about to receive a major shakeup.

Q. Why is there such complacency in the client/server world?

A. The major problem is that most decision makers in organizations, the high-up management, believe that PCs are toys and that they are not used in business for doing anything serious.

Q. Network managers certainly know otherwise, right?

A. The problem is they don't get listened to. The chief executive is unlikely to listen because for the last 10 years he has been listening to the PC children(stet) moaning about their technology. They have tuned out. ... This time it's not crying wolf.

They honestly believe that their trusty old mainframes actually run their businesses and that PCs are just provided to keep the employees happy. This is fundamentally wrong. We've seen reports that estimate up to 69 percent of Fortune 1000 companies use a PC as the main way of accessing their mainframe. If you fix the mainframe and forget about the PC, then you might as well go home.

Q. Don't you have a financial interest in hyping the problem?

A. I don't need to do this. I started a company (Network Managers) in 1990 with some friends and sold it to Microsoft in 1995. I don't need to work. ... I am about raising awareness. If I could jump up and down stark naked and it would get me front-page news for the Year 2000 problem, I would do it.

Before we developed any tools to address this problem, I established my reputation as a Year 2000 speaker and a recognized authority on Year 2000 issues for PCs. And it was only when we saw that nobody else was doing anything to help, that I set my engineers to task of developing tools.

Q. With so much of networked computing operating on Internet time these days, why won't natural hardware and software upgrades shield corporations from Y2K problems?

A. I don't believe that at all. For example, Office 97 has been released for five or six months now, and I have been into major corporations that are still thinking about whether or not to put in Office 95.

The software that most people are running on their desktop is the software that was installed when the PC was installed and most companies do not replace their PCs every two years, three years if you're lucky. An, in most corporations they don't throw them in the bin after three years, they just pass them down to somebody lower in the organization.

Q. What kind of network infrastructure problems do you anticipate?

A. There is no way, in my mind, that the Internet is not going to be affected by this.

In many cases, a router is very similar to a PC in hardware terms, and in software terms it just runs a very proprietary operating system and some very complicated software. But in essence a lot of these routers are just the same as PCs, so they've got similar types of problems both at a hardware level and at a software level.

I would expect (a non-compliant router) to actually fail to deliver packets, because it would incorrectly time-stamp the packets. It's not a problem with the actual routing protocol, more a problem with the way the routers interpret the protocol and the way that the different vendors have implemented it.

Novell is currently positioned as saying they are analyzing the situation and that they will get back to everybody at the end of 1997. This is a sensible approach, to assume they are not compliant, until they can prove whether or not they are.

It is my understanding from discussions with various people in the industry that it is (Novell's) intention to announce that NetWare 3.0, which is the most common network operating system in the world, will not be Year 2000 compliant and will not be fixed. If anybody wants to make it compliant they are going to have to upgrade up to NetWare 4.0 and probably to a new version of NetWare 4.0 that isn't even released yet.

A. What happens if you keep running NetWare 3 after 2000?

I suspect that your network will simply not work properly. Some of the problems we've seen with 3.11 NetWare are that it simply does not handle data after the 31st of December 1999. Any log-ins or any time stamp at the operating system level would not function properly and that might cause e-mail messages to be undeliverable or even have people not being able to log in to the network. Fairly substantial problems.

Q. What's wrong with 'Go to the upgrade' as an answer?

A. Only that it's going to be expensive. Almost inevitably the upgrade is going to require a bit more muscle on the part of the PC. You might need additional hardware as well.

Q. What advice would give network managers?

A. First they need to do a complete inventory of their whole hardware and software environment. For routers and hubs, they need to know what models they've got and what versions of firmware they're running in them. Then they need to get in touch with the manufacturers (to ask about compliance). Then they need to identify the mission-critical components; what do they need to fix in order that they can continue to operate?

At the end of the first stage, you need to raise board-level awareness of what this is all going to cost. At the end of the second phase, when you've figured out what you're going to fix, you need to go get the money from the board and actually get on with it.

Q. How does a company decide who to hire for help with Y2K?

A. All you can do is look around. The Year 2000 mailing list is a very good source of information. There are about 3000 or 4000 people on that list, most of them actually in the end user community, most of them are people who are in the middle of their projects. They have a lot of good advice about who is kosher and who isn't.

A lot of people in this industry, if it wasn't for the Year 2000 would be running UFO spotter societies. They're extremely committed, they're extremely passionate, but they're not very commercially aware.