DOWN FOR THE COUNT

by Wayne M. Krakau - Chicago Computer Guide, March 1993

It started with a phone call from a Novell Authorized LAN dealer asking for assistance in designing one of his first networks. He wanted us to work with him in interviewing his client, a law office, and planning a new network where only a couple of stand-alone PCs were already in use.

During the interview I discovered that many of the major decisions had already been made. The client's informal personal computer advisor (see my column from June of 1992 entitled "The Brother-In-Law Syndrome" for my opinion on amateur "advisors") had convinced the client to go with thin Ethernet using 8-bit cards in no-name (literally without a name plate of any kind) AT clones. Even the file server was to be an unsupportable clone of dubious heritage with only an 8-bit (XT class) NIC (network interface card). Multimate (YECHHHH!) was to be the main application program.

The senior partner in charge of the law firm's LAN project refused most of my advice. This included having a supportable (and Novell certified) machine as a file server, and equipping at least that machine with a 16-bit (AT class) NIC. It also included discussing the merits of UTP (unshielded twisted pair) wiring, especially in an office that wasn't laid out in a linear fashion - the only method supported by thin Ethernet. A requested check on local building codes regarding cable plants was also refused.

Considering the preeminence of WordPerfect in the legal market, I also questioned the choice of Multimate. It was invented to ease the transition from Wang dedicated word processing machines to personal computers and had no particular appeal as the choice for a firm that never had Wangs.

Later discussions with the LAN dealer were just as unproductive. He declined my offer of installation services. He also rejected my bid on the cabling. He told me that he had an arrangement with another firm that would do it for less than one-third of my price. When I cautioned him about the hazards of amateur wiring installations, he assured me that the firm he dealt with was a professional LAN cabling company. That ended our discussion.

A few weeks later I received another call from the same LAN dealer. He could not print over the network. He was also having difficulty getting Multimate and some other applications installed. I agreed to come to his client's office to fix the problems.

Using the P (Printer) and Q (Queue) commands on the main console of the Netware 286 server - in those days that name was still used - I discovered that no printers or queues had been defined. That was count one.

Prior to running Netgen, the installation program, to create a printer definition, I checked the underlying server configuration with the CONFIG command. That's when I realized that the server was configured to hold two NICs. Since this was a small, single-office LAN, this surprised me. A quick inspection of the back of the file server revealed only one card. The original installation had been blown. A second NIC had been defined where no actual card existed, causing a performance loss and providing a potential point for future system failure. That was count two.

The file server was the last machine in the daisy-chain of this thin Ethernet. Thin Ethernet demands that each end of the "chain" is terminated with a 50-ohm terminator, one of which (arbitrarily chosen) is grounded. The server had a home-made terminator consisting of a naked resistor poorly soldered to a short length of wire. It was a mess. That was count three.

Since this end of the network was ungrounded, I went to the farthest office to check the last machine in the original plan. It also had a bizarre terminator. That was count four.

Neither end of the LAN was grounded. That was count five.

Now, it was back to the file server to trace the rest of the network. The segment from the file server to the first workstation looked just fine, but the next segment was lying on top of an under-desk heating unit. The immediate problem was the electrical interference, with the long term problem being the heat induced degeneration of the cable. Naturally, I removed the heating unit. That was count six.

I decided to examine the first workstation. It booted off a floppy disk and automatically logged in to the network. The distinctive sound of a hard disk spinning up to speed emanated from the machine. Sure enough, it contained a perfectly good, bootable hard disk. The floppy boot disk was worse than useless - it slowed the boot process as well as access to DOS commands. That was count seven.

Based on a report that some of the workstations would randomly lose access to the network, I ran my favorite LAN diagnostic software, TXD, from Thomas-Conrad Corporation (Austin, TX). It uses Netware's built-in error detection capability that normal programs ignore. Upon initialization, it scans the network (or internetwork, if applicable) for all servers and workstations. In this scan, it couldn't find the workstation on which it was running! This was not a good sign.

Next month, the conclusion of this network horror story.

                                    1993, Wayne M. Krakau