DOWN TO THE WIRE

by Wayne M. Krakau - Chicago Computer Guide, June 1994 - NewsWare, March 1996

Due to the unprecedented response to my last two columns, "Hot Wired," parts one and two, I decided that a follow-up was in order. The general theme for many of the inquiries I received was "I want to fudge the standards in the following way and have been advised by ________ (fill in the blank, usually with "cabling company", "consultant", or some variant of "reseller") that it will work -- can I get away with it?" That's a scary thought.

In one typical example, a 4Mb (megabit) token ring network was being upgraded to 16Mb. The exiting Level 3 (10MHz rated) cable was augmented by replacing all external portions (jumpers) in the wiring closet and at the workstations with Level 5 (100MHz rated) with the expectation that the cable plant as a whole would now be able to support 16Mb token ring.

I suspect that the math required to believe in the viability of this "upgraded" system is some kind of averaging: (Level 3 + Level 5)2 = Level 4 -- right? WRONG! Level 5 added to a Level 3 cable plant is usually worse than having all Level 3.

Each spot at which two different kinds of cable meet creates a potential point of distortion. Multiply that times the number of individual lines used and then again by two to account for one jumper at each end of the internal lines. Even if this effect wasn't a consideration, you have to remember that cables work on the weakest-link principle. Even one bad or inappropriately rated segment can wreak havoc on an entire network.

I think I may have inadvertently set myself up as some sort of father-confessor. People seem to be asking me for a special dispensation so they can use invalid cable plants. Sorry, I can't give one. The Laws of Physics keep getting in the way. Feel free to invoke the deity of your choice, and be sure to get back to me if you are successful. That would be in the same category as a weeping icon.

Here's the way to avoid messes like this one. First, deal with cablers who are familiar with the rules. This doesn't mean that they necessarily have memorized every rule, just that they know the general parameters well enough to recognize possible deviation and aren't afraid to look up the details (especially for the potentially complicated calculations required for token ring!).

Also, they must have the ethics ("Have pulpit, will preach", is my motto) to turn down cable installations that are inherently invalid. If you find discrepancies via other sources, and a cabler is still willing to install the invalid cable plant, look elsewhere!

Don't allow the cabler to get away with merely testing for continuity. They must use appropriate testing devices. While there is technically no such thing as a true "certifying" device, there are several instruments that can come as close as possible to this ideal. A bonus for this type of product is that most of them can print a cross-reference report listing the individual cables that were tested along with the details of the test results.

The final qualification of the cabler company is the willingness to back their work. Will the company send qualified technicians to fix problems? Will they reimburse your company for expenses due to having a LAN person track down mysterious errors that turn out to be cabling problems? This is the final and most important characteristic to look for. After finding errors in many existing cable plants over the years, I have found that most cablers won't support their work, even in the face of overwhelming evidence of a defective system.

The King is Dead, Long Live the King

The mournful sound of Taps wafts over the fields. A riderless horse with reversed boots in its stirrups passes at a slow walk. Artillery is aligned for a final salute.

No, I am not talking about deceased ex-presidents. I am talking about the recently announced demise of Netware 2.2. Novell has finally pulled the plug on the last direct descendant of what was originally known as Netware 286, itself a derivative of the old Netware 86 and in turn Netware 68. (Note that the digits were added after the fact for differentiation of each type from its successors.)

Like the recently expired ex-president, Netware 2.2's reign was a controversial one. Novell released it in the aftermath of a wave of protests about a proposed freeze on development on what was then known as the Netware 286 line. Novell succumbed to an immense backlash. People resented being "forced" to upgrade to Netware 3.11. That's like being forced to go from a hand-cranked engine to an electric starter.

While it carried on the legacy of its forebears (Netware 2.15c being its immediate predecessor), it really was a dead-end track on the evolutionary trail of network operating systems. Released at a time when it was generally acknowledged that the Intel's 80286 was dead, Netware 2.2 ended up being a boat to nowhere.

Unscrupulous and unknowledgeable resellers were happy to sell an operating system with a proven record of high installation and maintenance costs. (To this day, Novell's competitors continue to hold up this reputation as evidence that prospective customers should consider their operating systems, or just to get people to move to peer-to-peer systems. They neglect to tell people that a simple Netware 3.x LAN is actually easier to install than a peer-to-peer system for an installer experienced in both.) The resellers had the bonus of knowing that their customers would eventually have to upgrade their LANs when the artificially resurrected 286 line ultimately became extinct. Depending on the number of 2.2 boxes left on resellers' shelves, this opportunity will soon be eliminated.

Novell has taken a carrot and stick approach toward motivating the remaining Netware 2.x users (over one-third of all Netware customers, the last I heard) to upgrade. The carrot is a discount on upgrades to either 3.12 for 4.01 Netware with the promise of a free upgrade to the upcoming 4.1 for those who choose 4.01. Even 3.11 customers get the discount when upgrading. The stick is the threat of a gradual loss of support for older systems.

An interesting side note to this is that I just found out about an unpublicized upgrade opportunity for owners of older versions of Netware at least back to 2.0a, and probably including 2.0, too. Even though upgrades from these versions have long been unavailable through resellers, I recently found that they can be obtained at bargain prices directly from Novell. Just call 1-800-NETWARE and walk your way through the menu system to get the sales department involved with upgrades.

Since I have had multiple conversations with Novell personnel in other departments (mostly the reseller group) during which I was told that older versions of Netware were not upgradeable, I was quite surprised to hear of this option. I suspect that the impending doom of Netware 2.2 might have some effect on the availability of these old-system upgrades, so if you are interested, act now!


1994, Wayne M. Krakau