Windows manager: Watching for bugs in Windows 98

Windows manager: Watching for bugs in Windows 98

Regardless of the outcome of the Microsoft antitrust case, one thing is certain: some product called "Windows 98" will eventually be available to the public. Because that is the case, we should start considering now what problems, if any, will come with an upgrade from Windows 95 to Windows 98.

I can't predict every glitch that will occur when people with a wide variety of systems start using Windows 98 in real life. I invite readers to start sending in any secrets you learn when using Windows 98.

What we know from the past is that some of the worst side effects of upgrading to new Windows versions were not mere bugs but deliberate decisions made by Microsoft. Because these "gotchas" occurred more than three years ago, most people don't remember them, so it's worth reviewing them now.

* System conflicts. When companies upgraded from Windows 3.0 to late beta versions of Windows 3.1 in 1992, those using a DOS version other than Microsoft's MS-DOS saw the message "Non-fatal error detected: error #2726", and Windows halted. The message instructed users to call Microsoft's beta support program. When people called, they were told Microsoft would not provide technical support for Windows if a non-Microsoft version of DOS was used.

This error message caused many companies and PC manufacturers to stop buying DR-DOS, a competing operating system that was available from Digital Research, which was a major software company at the time. If Windows wouldn't run with DR-DOS, that was enough to persuade most buyers not to use it.

DR-DOS did nothing to cause the error. Microsoft had simply written "some MS-DOS detection code", as one Microsoft executive described it at the time. I wrote about this in November 1993, thanks to research by Andrew Schulman, author of Undocumented Windows (Addison-Wesley).

* The Internet shuffle. The upgrade from Windows 3.1 to Windows 95 in August 1995 caused another serious headache. Prior to Windows 95, PC users ran a variety of software to access Internet services. Key to these services was a file called WINSOCK.DLL. This file makes it possible to connect to remote computers using IP.

Installing Microsoft's TCP/IP support or Microsoft's Internet Explorer 3.0, however, replaced competing versions of WINSOCK DLL with a version that worked only with the new Microsoft Network. This disabled competing services, such as CompuServe's NetLauncher, Spry's Mosaic, and others.

Many users of these products tried to reinstall their applications to make them work again. But the next time that Win 95 restarted, it detected the competing WINSOCK.DLL and replaced it again with its own proprietary version. I wrote about this in 1995, and Internet services soon found ways around it, but it created serious problems for Microsoft's competitors when the Net was young.

If the Justice Department can keep Microsoft from using Windows 98 to play dirty tricks such as these on competitors, it will have earned its keep. When you first install Windows 98, send me e-mail with the subject "upgrade secrets" and let me know what you find.

Brian Livingston is the co-author of several best selling Windows books, including the most recent Windows 95 Secrets (IDG Books). Send tips to He regrets that he cannot answer individual managerTackling IP addresses I am responsible for maintaining a Windows NT 4.0 network. Currently, my network is a Class C IP based on with a subnet mask of My Internet router and default gateway is 0.65, and our DNS server is yyy.yyy.yyy.2 based with our ISP and accessed through a 256Kbps frame-relay circuit.

We have recently added two additional networks with routers on our LAN at 0.120 and 0.121. These routers are connected by additional frame-relay circuits with remote routers and 192 .100.100.1. These devices have PCs using and respectively, using multigateways of 120 and 65. We have placed static routes in each remote router for the 0.65 router and with 0.120 and 0.121 as the next hops.

We can PING the devices and they can PING the Internet router at 0.65. However, they cannot access the Internet using the 0.65 router. The 0.65 Internet router does not have RIP capabilities; however, we have built static routes for 0.192 and 0.194. - Doug SeaBased on the network numbers you have specified, I'm assuming you picked them arbitrarily and they are not registered to you. So when PCs on those networks send packets out to the Net, responses will be routed back to the real and networks.

Routers on the Internet keep tables of how to get to the many thousands of networks that are connected. For instance, when an internal client requests a Web page, that client's TCP SYN request probably heads out just fine. But when the Web server sends back the SYN/ACK agreeing to the connection, it is sent to the real networks that use those addresses.

You've got two options. The most straightforward would be to get two additional networks assigned to you from your ISP. Because you're already subnetted with 0.192, it would make sense to get the entire network assigned to you and, if you can use subnet zero, break it into three networks with the 0.192 mask that you're already using. If that isn't an option, you must move your new networks to unregistered addresses. Addresses that begin with 10. or 192.168. have been set aside for IP networks that are not connected to the Net. Then you can use either a firewall or router that does Network Address Translation, or a proxy server to connect your unregistered networks to the Net.

You could go on using the 0.194 and 0.192 network numbers you've created, but nobody in your network will be able to get to the real sites that use those addresses because packets that are intended for those networks will be sent to your own internal networks. Troubleshooting that kind of problem can be a real pain, so I wouldn't recommend that route.

Pentium Pros

In your May 27 column you wrote that because the Pentium II is based on the Pentium Pro it does not run DOS applications as well as a Pentium. We have a large network running a DOS-based application and need to upgrade some of the machines. Would a 266MHz Pentium run faster (and save us money) than a 266MHz Pentium II? Would this hold true even if we ran the DOS application in a Windows 95 DOS window? - Kevin Anderson As long as your DOS application is an eight- or 16-bit application, odds are you'll find better performance with the 266MHz Pentium machine than with a comparable Pentium II. If you're planning to buy machines and keep them for a while (in which case they may be used with Win 95, Win 98, or NT), it's probably a better idea to go with the Pentium II and live with the slight performance hit. The machines will still be faster than anything imagined when your DOS application was written.

Laura Wonnacott is drinking margaritas on a beach somewhere. This week's column is by Test Centre manager Brooks Talley, who has worked with corporate networks for 12 years. Send comments to

Follow Us

Join the newsletter!

Error: Please check your email address.
Show Comments