First Impressions

Simson L. Garfinkel

The week before NeXTWORLD Expo, I spent two days trying to install NEXTSTEP on a Gateway 2000 with an Adaptec SCSI controller. The problem: My CD-ROM drive had been set to SCSI target ID No. 6 instead of No. 1. According to NeXT, Adaptec requires that each SCSI device have a sequential target address, without any gaps or empty addresses.

The only problem with this explanation is that it doesn't have to be true. While Adaptec's ROM may have a brain-dead search algorithm, NEXTSTEP doesn't have to emulate that behavior. If NeXT had written its Adaptec driver to reprogram the board and then scan every SCSI address, I would never have had any difficulty getting NEXTSTEP installed on the Gateway.

The week following Expo, I tried to get NEXTSTEP to install on a Dell DGX with a DPT SCSI controller. After three days and numerous telephone calls to NeXT, I finally tracked down the problem: NEXTSTEP doesn't work with DPT's default configuration. To get the operating system installed, I had to run DPT's configuration program and change the board from its "primary" to its "secondary" configuration.

Once again, it's a problem that could have been avoided with a better NEXTSTEP driver. Instead of requiring that the DPT board be in its secondary configuration, NEXTSTEP should work with either. (NeXT claims that people don't use DPT's primary configuration because it conflicts with IDE drives, but most people who purchase a DPT SCSI board probably aren't going to have IDE drives.)

These are not isolated incidents. Lots of people even gurus inside NeXT are having problems getting 3.1 running on white hardware.

I tried to install NEXTSTEP on a third system. Each time, the system halted with a cryptic error message: "vnode_pagein error ... Exception #3 raised, press c to continue." It sure seemed like a bug in the operating system to me, so I called NeXT to report it. A few hours later, I discovered that the fault was my own: improper SCSI bus termination. But the whole thing could have been avoided if NEXTSTEP's installation procedure included a self-test of the CD-ROM drive and printed an error message if the driver wasn't up to spec.

Don't get me wrong NEXTSTEP 3.1 is the most reliable version of the operating system that NeXT has ever produced. But NeXT has to realize that first impressions count. The problem isn't bugs with the software; it's an unwillingness on the part of the software to be more forgiving of typical human errors. And in the wild and wacky world of PC hardware, software almost always gets blamed, when the real fault lies with configuration snafus or flaky hardware.

NeXT must improve NEXTSTEP's installation procedure. NEXTSTEP has to sense the computer's configuration and accommodate it. NEXTSTEP must perform rudimentary self-tests on the computer's hardware and print diagnostic messages that can be understood by mere mortals. The installation and boot-up programs should be modified to incorporate "watchdog timers" so that, if the system crashes, the watchdog utility at least tells the user and gives some indication of what went wrong. Otherwise, many people evaluating NEXTSTEP might give up on the operating environment before they even get it up and running.

Simson L. Garfinkel is a NeXTWORLD senior editor.