An Unabashedly Biased View of the Passing of a Network Titan

An IDC report from December 2004 contains some interesting figures for NetWare, the former king of the server room. From an installed base of 2,336,000 units this year, IDC anticipates a drop to 1,408,000 by 2008 -- with a similar story for new license shipments. Extrapolation and statistics have never been my forte, but suffice to say, if IDC is spot-on, then around 2019, things will start to get pretty asymptotic for this once mighty leviathan. It may by then be forgotten that there was a time when the world of the server was almost exclusively red, had teeth and spoke in a languid Midwestern drawl.

Anyone working in IT in the mid '90s will remember the bold (and wholly truthful) claims that Novell Inc.'s NetWare powered 65% of the world's servers. Competitors at the time such as Banyan Vines (remember them?) could only watch, mouths agape, as the red-boxed beast of Utah ran rampant over their sales figures. I was working at the time in London, where full-page ads appeared in mainstream papers like The Times for jobs with dizzyingly attractive employers such as big U.S. broker Lazard Bros. They asked questions like, "Could you Novell in the City?" and offered salary levels best described by borrowing a line from Big Daddy Joe Cabot in Reservoir Dogs: "Juicy junior, real juicy." Twelve years later, I'm still nowhere close!

Sadly, nowadays all a Certified NetWare Engineer qualification and the ability to repair a NetWare volume will earn you is a wry, pitying smile from the 21-year-old, pigtailed Debian freak whose invitation to interview was delivered to you by Short Messaging Service (you had to get someone else to decode the "txtspk") and who deep down you know was doing you a favor by even looking at your resume. You leave the building wondering what the term modprobe means and why everyone was sporting penguin lapel badges.

Where Did It All Start to Go Wrong?

Leafing through my old copies of the NetWare Technical Journal, an early and very high-quality attempt to spread the gospel for a quaint new craze called "disk sharing," one finds the following gem of an ad in the January 1989 issue:

NetWare Core Protocol Documentation of NetWare's client-server protocol. $10,000

Not exactly a bargain. Admittedly, 1989 could well be considered the zenith of the proprietary software gold rush, but as a means of enhancing exclusivity and enticing only the largest and most profitable of organizations to develop to your platform, a $10,000 price tag for a few hundred pages of A4 is a pretty good start. If you're into ensuring long-term survival for a nascent operating system of as-yet-unacknowledged speed and power, however, it's the equivalent of the hangman setting up the gallows. Couple extortionate prices with developer certification programs that forced coders into ever more supine positions of worship, and you start to wonder just why the open-source methodology took as long as it did to gain traction.

Getting hold of the docs wasn't the only obstacle, however. The early versions of the NetWare operating system forced developers to link their processes into the kernel and then reload it. For committed speed freaks, this was a hassle too far. By the time NetWare 3 came around, the world was introduced to the concept of the loadable module -- quite some time, incidentally, before the now-ubiquitous Dynamic Link Library. Once you got developing to this innovative new architecture, breathtaking speed -- back in the days when network operating system performance mattered -- was yours for the taking. Problem was, not many folks got it. NetWare 3 had several different memory pools, and as a developer, you were expected to choose the right one for the job at hand. Speaking as someone who used to debug NetWare core dumps, I can tell you that many developers simply couldn't be bothered and instead of juggling their own pools simply ended up wading through each others'.

NetWare got an undeserved reputation as being hard to code on. One of the main reasons NetWare has such high performance is its nonpreemptive nature. When your operating system has to mollycoddle each and every process it hosts to ensure that all get a fair share of the CPU, then you have overhead. Overhead equals bloated code, which equals reduced performance. In the NetWare world, developers had to seek out the creeks and gullies in their code where they could safely sit back, put their feet up and just chill for a millisecond or two. Problem is, coders are lazy. Nine times out of 10, they'll take the easy option, which meant minimal to zero yielding, which often led to freezes, hangs and a whole host of other nasties. Microsoft Corp.'s alternative was a comfy environment where, for the cost of performance, you could behave as badly as you liked knowing that Big Brother kernel32.dll would look after you if you got too far out of line.

The Sleeping Giant Awakens

With the later NetWare Versions 4 and 5, things were looking a lot better. There was simplified memory, symmetrical multiprocessing support and, with NetWare 5, a truly blinding Java virtual machine and a fancy new file system. By this time, however, Microsoft had its act together. The abomination that was Windows NT 3.51 was replaced by something marginally less scary, NT 4.0 -- the bold (and wholly untruthful) claim being that it was New Technology. The techies of the world weren't fooled -- even today in Windows servers you can dredge up obscure error messages with the text "LAN Manager" screaming out at you. The only thing new about NT was its Swiss-cheese approach to security.

Alas, two factors combined to really erode NetWare's market share more than anything I've already mentioned: Microsoft's impressive marketing machine and the wholesale absconding of purchasing power from the geeks to the boardroom. The two are closely interlinked. If you're the CEO of a successful company, constantly bombarded with slick, professional messages at airports and hotels, from radio and TV, and even on airplanes, telling you that you need to take more control of your IT, then it's only natural you start to ask just what it is those guys in the basement are up to. And let's face it: What company boss doesn't like to add another territory to his empire now and again? Once you've got the CEO's, CIO's and CFO's attention, simply crank up the volume -- and Microsoft's marketing volume sure does go all the way up to 11 -- then sit back and watch the wins roll in.

And the Future?

It's tempting to round this article off with the cliche "but there's a new kid in town." The thing is, in the server world, Linux and open-source aren't new. They're tried and tested and have hit their competitors where it hurts. Microsoft, Sun Microsystems Inc. and a whole host of others have recently made conciliatory gestures toward the Linux/x86/open-source triumvirate for very good reasons: Linux, because these biggies of the IT market can see what has been crystal clear to Linux customers for some time -- total cost of ownership is way down; x86, because it's cheap and, well, let's face it, anyone still running Sun hardware in 10 years' time will need their own coal-fired power station to supply the juice; open-source, because charging someone thousands of bucks for the privilege of coding to your platform is quite frankly an affront when, for the same money, you can get a high school in Mumbai to deliver you a shiny new kernel.

I actually happen not to believe the doom-mongers of IDC who, although they may have got the trending accurate, omit to envisage some sort of future role for NetWare. Make no mistake, NetWare isn't going away. The glory days may be over, but Open Enterprise Server with NetWare is a seriously compelling offering -- the niche it will eventually find is, dear reader, a subject for another time.

Paul Coletti is an Edinburgh-based IT consultant and blogger who has worked in IT for 16 years at IBM, Novell and other companies.

Copyright © 2005 IDG Communications, Inc.

7 inconvenient truths about the hybrid work trend
Shop Tech Products at Amazon