The Hatchet Job

Sponsored by


The news media in general, and the computer industry media in particular, are often believed by the general public to be relatively unbiased in what they publish. In other words, despite the fact that media personalities are "real people" just like you and me, with feelings, beliefs, and biases, we somehow believe that (for the most part) a journalist would never intentionally botch a story. We expect that someone who makes a living telling people what is going on in society does little more than that -- tell people what is happening. They may not tell the whole story, or they may emphasize one aspect of a story in a slanted manner, but we have a degree of trust that the overall message is relatively accurate. We don't live in a closed, totalitarian society, so obviously the news media are not totally falsifying the news, right?

However, occasionally an article is published that is so biased, so disingenuous, so phony as to challenge our assumption of relative fairness and decency. If this blatant falsehood is carried on by a particular publication as part of a long-term trend, we must face that facts and reassess our willingness to give that publication and its staff the "benefit of the doubt." I will now dissect an article published in InfoWorld magazine's online edition (click HERE). This article leaves no room for doubt about the level of extreme bias against superior computer technologies possessed by certain members of the InfoWorld staff.


The reviewer is Tom Yager, a regular InfoWorld contributor who operates a private testing lab in Texas, according to the article. Visiting his site (click HERE) shows that he is Senior Contributing Editor and Columnist for Windows NT Systems magazine. What kind of bias do you think this paid Windows NT writer would bring to an OS/2 Warp review? What kind of trade magazine would select this person as an acceptable OS/2 reviewer?


The review article discusses a network server software package. The purpose of a network server is to provide file transfer, file sharing, storage, printer sharing, and the sharing and management of other resources on a network of personal computers. The key elements of a server's performance are first and foremost, reliability; second, scalability. This is the ability to add new users and share additional resources, and perform ever-greater number of transactions per unit time without suffering significant performance or reliability degradation. A server is usually more or less "invisible" in that the users will not even be aware of its existence if it functions properly. Similarly, technical personnel want and need a "set and forget" server that can be installed once and thereafter requires little or no interaction except when changing the system configuration (adding or deleting users or reallocating shared resources).

Of secondary importance is ease of system management and reconfiguration. It is expected that in most large installations there will seldom be a need to perform wholesale changes in system architecture and resource allocation. Typically the system management involves a handful of utilities and perhaps a few command-line scripts. This is not an activity that requires a high degree of user intervention, and hence a complex or fancy user environment is not required (and indeed may be an unnecessary distraction).

Finally, it is desireable to have good system documentation and relative ease of installation and configuration. However, a server that functions correctly will not need to have constant attention or re-installation in the long term. Therefore, these factors should be of relatively minor importance to the overall perspective of a network server review.


The product under review is IBM OS/2 Warp Server 5.0. This product has a sterling record of reliability on par with Unix systems. The product is capable of serving thousands of terminals on a single server. The innovative user interface has long been recognized as a the finest example of object-oriented technology with a user-friendly appeal. While not among the most popular platforms, the product has nonetheless gained a loyal following and in some cases (such as the banking industry) it is the clear leader as a state-of-the-art platform.


The article itself is now quoted in Yellow, interspersed with an analysis in White.

IN THE BATTLE for the PC server market, every player needs to set its operating system apart from the others. Microsoft throws mountains of functionality into Windows NT and gives its users massive free updates. Novell NetWare has good performance, leading-edge directory services, and a loyal installed user base. Linux is almost free and a buzzword as well as an OS. And IBM's latest release of OS/2 is expensive.

The reviewer begins his work by summarizing his viewpoint on the essential features of each operating system. However, he immediately betrays his bias by refusing to cite any performance or functionality traits for OS/2. He mentions the "massive free updates" for WindowsNT, but neglects to state that OS/2 provides regular free fixpacks that add significant enhancements. He cites Novell's good performance and loyal user base, but neglects to mention that these are also well-known aspects of the OS/2 platform as well. He then sets up OS/2 by citing Linux as "free" and a "buzzword," two things that OS/2 obviously is not. The intentional implication is that OS/2 does not have *any* of the previously listed benefits, which is of course untrue.

Let's try this out for size: "Chicken is delicious and has a strong cultural attachment. Hamburgers are everywhere. And steak is expensive." See how biased -- and even stupid -- this argument sounds? The fundamental difference between steak and mass-market food is taste; you pay more for a better-quality meal. Similarly, OS/2 costs more because it has superior performance.

To give IBM its due, OS/2 Warp Server for e-business is the most powerful OS/2 ever released. It is Warp Server 4 with accumulated fixes and a handful of new features: a journaling file system, year-2000 and euro currency support, a Unix Network File System file-sharing client, four-CPU multiprocessor support, a Java-enabled Web server, and a logical volume manager that permits some changes to the file system configuration without rebooting.

Well, that sounds like a nice laundry list, but it breezes through the list without explaining anything. Just what is "a journaling file system," and why should I want one? What about OS/2's ability to expand beyond just four CPUs for SMP support?

Like Windows NT, this version of OS/2 crams a lot of functionality into one bundle. But unlike with NT, you pay for that power by enduring an outdated user interface, scattered and concealed features, abominable documentation, and a tiny set of third-party applications.

Now the attack begins in earnest. Calling WorkPlace Shell an "outdated user interface" is about as insightful as calling the Internet "just a bunch of wires." Just exactly how does a user interface become "outdated?" Does it have wrinkles? Does it wobble when it walks? And besides that, why would a network manager want to have a user interface change every few years, particularly if the original interface does the job effectively? Who needs constant retraining expenses? And just how much time does the network engineer spend interacting directly with the GUI, anyway?

While the features may be "scattered" in different folders, being "concealed" has a lot more to do with knowing where to look than it does with any supposed attempted to "hide" something. Certainly a universal management console would be helpful, so this may be a point against the Warp Server. Similarly, the documentation is classic IBM -- either cryptic or nonexistent. However, the sentence switches gears once again and focuses on a supposedly "tiny" set of applications. Since when does a network server need thousands of applications? Its job is to provide system management functions, not gameboy features for bored back-office nerds. Besides, the fact that Warp Server's high-performance Java engine makes all Java programs instant third-party native applications is lost on the reviewer, who fails to even mention the embedded Java API and its industry-leading performance.

Thus one sentence combines a personal bias against the GUI, two valid complaints, and a misleading statement that ignores the vast and growing number of Java applications in existence.

Last December I looked at the beta release before IBM had announced pricing, and I held out hope that OS/2 would be priced to compete with NT. But OS/2 Warp Server for e-business is more expensive than Windows NT Server 4.0, turning out to be no competition at all. Unless you already run an OS/2 shop and do not want to jump ship, you will find few reasons to deploy this OS/2.

This reviewer's obsession with pricing betrays his ignorance of relative server capabilities. If price is really so important, why not go out and buy a copy of Multiuser DOS and use it as a server package? There is more to a server O.S. than simply comparing features and prices.

For example, how many workstations can reliably run on a single copy of NT4 Server? How many for WarpServer? How many times per month will the NT crash, versus the OS/2? How badly will NT's performance drop off after adding 100 more users, versus OS/2's performance scaling? To put the price argument in perspective, how many copies of the NT4 Server (as well as associated hardware) will you need to buy to match the performance, the reliability, and the scalability of just one OS/2 WarpServer?

This review is a perfect example of being "penny wise and pound foolish." No doubt a large percentage of the two trillion dollars spent on information technology in the last few years by U.S. companies is another such example. I suppose that kind of stupidity is one reason why WindowsNT is still used, despite the fact that OS/2 WarpServer is actually a less-expensive solution. That's right, if you count the cost of doing business instead of just the price on the label, OS/2 is far cheaper. The real problem here is that the reviewer is fooled by the unrealistically low entry cost of the Windows platform, and fails to recognize that a network is more than just the sum of its parts.

I tested OS/2 Warp Server for e- business on a PC server with a Tyan motherboard, dual Pentium II CPUs, 192MB of RAM, and a 9GB Seagate Barracuda ultrawide SCSI primary hard drive.

With how many workstations attached? Zero? That explains a lot, doesn't it?

You boot the OS/2 install from a CD, so if you want bootable floppy disks, you must make them yourself. During the installation, you must constantly swap the installation and Server Pak CDs because the boot CD contains no installable software. The installer reboots your server several times, often without warning.

When OS/2 came on diskettes, reviewers balked. Now a boot CD option has been added, and this reviewer whines and complains. Considering how easy it is to make a couple of diskettes, this is a trivial complaint that has no business in a serious product review. Now the need to swap install and server CDs is indeed a negative point, but why mention the reboots being "often without warning?" What kind of warning would you like, a dancing elephant? What will you do when the warning occurs, duck and hide under your desk?

The OS/2 desktop is still painted by Presentation Manager, a Windows 3.1 contemporary.

The writer is in error by failing to tell the whole story. The WorkPlace Shell (WPS) GUI of OS/2 was smartly updated with the release of OS/2 Warp 4.0 to include the Warpcenter, 3-D icons, TrueType font support, and more. Since OS/2 Warp 4.0 was released in September of 1996, this makes the OS/2 GUI two months newer than the Windows NT 4.0 GUI. In other words, the OS/2 GUI is in reality a "Windows NT 4.0 contemporary." Since the OS/2 GUI is newer, does this imply that the Windows NT GUI is "outdated," then?

Besides, since when is keeping a superior GUI a negative point? This argument is completely empty-headed. Let's see how this sounds with more everyday items: "Nobody should still be driving a 1968 Mustang, a VW contemporary." "Nobody should still listen to the Beatles, a Herman's Hermits contemporary." "The 1985 Chicago Bears were losers -- they were Tampa Bay Buccaneers contemporaries." Placing a time limit on great products falsely implies that anything new is automatically better than anything old. This is a major cultural fallacy of the early twenty-first century. It is too bad the reviewer has fallen into that trap himself.

The default desktop's root window icons and taskbar menus look haphazardly laid out by a dreadfully disorganized user. Essential functions are buried in submenus, and administrative tools and utilities are scattered and concealed in the pages of anachronistic on-screen notebooks. It is a GUI OS/2 fans have learned to love, but a hard taste to acquire if you are accustomed to Windows 98 or Linux's Gnome.

Hello, wake up in there! Have you ever heard of right-click, arrange? OS/2 allows you to make a fresh folder with shadows for just the tools you use regularly, and keep the more obscure stuff hidden out of sight. That is actually a nice advantage of a true object-oriented interface. And what is this garbage about "anachronistic" notebooks? Since when did notebooks become obsolete? How many million units of notebooks are sold each month in this country? Would you prefer a pop-up Elvis impersonator for every menu choice?

The idea that the OS/2 GUI is a "hard taste to acquire" is totally backwards reasoning. Let's try that argument with foodstuffs: "Sirloin steak is a hard taste to acquire when you've become accustomed to McDonald's or KFC." Do you see how stupid the review looks when its bias is exposed?

If you have the patience (or experience) to find your way around, you will find some worthwhile elements. In a mixed LAN with shared NT servers, OS/2 will replicate the user list and let you set up a single-login environment. The new Journaled File System improves reliability and speeds recovery after a power or software failure. OS/2 Warp Server for e-business plays well as a peer in an NT LAN.

IBM touts online volume management as a key new feature of this OS. The Java-based Logical Volume Manager (LVM) brings less to the party than is needed--you can't format a volume or change its file system type from within LVM, and it does not support software RAID, as do NT and Linux. But you can extend a journaled volume without rebooting, which worked in my testing. The LVM tacked the new space onto the end of the existing volume, while retaining the existing volume's contents.

The one killer app in the package is IBM's WebSphere Applications Server. WebSphere equips Web developers with robust server-side Java programming support via servlets and Java Server Pages. CORBA object broker support is built into other versions of WebSphere, but sadly, not this version. Java outperforms script, potentially making for more scalable Web applications, but the missing CORBA support leaves OS/2 with nothing to compare to Microsoft's Component Object Model technology.

FUD alert! Notice how casually the reviewer slipped from "WebSphere for OS/2 lacks CORBA support" to "OS/2 lacks CORBA support?" Certainly IBM should get in gear with a fully-functioning WebSphere, but there is no need to use that point to FUD the whole platform.

The OS/2 Warp Server for e-business literature winks at a comprehensive Java environment, promising the "OS/2 Warp Developer's Kit, Java Edition." This turns out to be the Sun Java Developer's Kit, which is a free download, some online documents, and an IBM just-in-time Java compiler. An editor touted as the "Java editor" is passable but offers no special Java-related features such as syntax highlighting or statement completion.

IBM also supplies a folder full of links to Sun Java examples, copied to your hard drive during installation. Many of these did not execute during my testing. This may be a good place for an experienced Java jock to work out, but it certainly isn't the best place to start.

Agreed, most IBM examples are more in the vein of "show and tell." But nowhere in this entire review was the issue of relative Java performance on different Server platforms even mentioned. Perhaps it was too embarassing to admit that OS/2 continues to drive the industry forward in terms of Java performance?

The Domino Go Web server, an adapted edition of the Unix/Linux Apache server, does a fine job with basic Web content. Domino Go improves on Apache with a Web-based administrative interface. Compared to Netscape's administrative front end, Domino Go's GUI is fractured and disorganized. Basic configuration requires lots of tiny steps. This is a far cry from Windows NT's IIS, which groups all of your Web parameters together whether you are administering it locally or from the Web.

Once again, the parts you are given can be rearranged at will using that OS/2 GUI. You know, the GUI with the "hard to acquire taste?" Windows had better provide you with some kind of pre-arranged gizmo, because working with goofy "shortcuts" can be a nightmare compared to OS/2's elegant icon shadowing function.

The Web server is quite functional and includes Secure Sockets Layer and basic (Web, FTP, and Gopher protocols only) proxy support. I am a fan of the Apache server, and I advise you to forego Lotus' tedious Web administration interface and edit the server's configuration files directly. But don't get too attached: IBM is discontinuing Domino Go Web Server. Support for it is scheduled to be discontinued in 2001.

Telling half the story is as good as a lie in some cases. IBM pre-announces discontinuance of products and then periodically extends the support period. This has been true at IBM for years. This still occurs with OS/2 and a number of other IBM products. This is the opposite of Microsoft's pre-announcing delivery and then failing to deliver. Besides, why not ask IBM what they are planning to replace the Domino Go with? That would actually be some new information, something this reviewer apparently has little of.

WebSphere holds tremendous promise for hosting server-side Java Web applications, but it is not enough to carry OS/2 Warp Server for e-business. No combination of its features could make this costly OS appeal to anyone not looking to upgrade from a prior release.

"Anyone?" Once again, the reviewer attacks cost without explaining the performance benefits -- because he obviously failed to test the Server's performance. A server review that ignores performance is like a basketball box score that fails to list the points scored. You know who the teams were, and you know the names of the players, but you have no idea whether anybody did anything or not. This review is so useless as to be a travesty.

If you like WebSphere, run it on IBM's AIX, Windows NT, or one of its supported Unix variants. That will cost you less, and you'll be using IBM's advanced server-side Java on an operating system with a future.

The reviewer cannot help himself. He has to take a parting shot with the boring, worn-out rumor that FUDs the future of OS/2. Yes, this is yet another FUDmeister review that says a lot but tells us nothing of real value.


A bad product review is not a crime, at least not when it occurs in isolation. But a bad product review when published by a major trade magazine also reflects badly on the editors and the managers that approved its publication. How could an ostensibly responsible computer magazine publish a review of a network server that was never networked and never served anything? Knowing that IBM is targeting OS/2 shops, why not have a regular OS/2 user (instead of a professional Windows NT writer) perform the product review? Would selecting an OS/2 developer to do a product review of a Windows98 release be considered "unbiased journalism?"

This is the same magazine, InfoWorld, which withdrew its Product of the Year award in 1997 because the "wrong platform" won it -- OS/2 Warp. An editor at this magazine claimed that OS/2 "zealots" had "stuffed the ballot box," yet the entire voting population was only some 400 people. Does 400 votes out of 300,000 readers sound like ballot stuffing, or rather a poorly-conducted poll? The magazine chose to use trumped-up charges of tomfoolery in order to cover its own bad work. The fact that OS/2 won the poll for a fourth straight year angered the people who have staked their careers on error-prone, dead-end platforms like Windows. Another incident shows the emotional investment of top InfoWorld management in Windows: after he publicly refuted Microsoft's wimpy excuses in the DOJ trial, InfoWorld fired "Help Desk" guru Brett Glass.

This is also the magazine that published a series of anti-OS/2 articles, including one by Ed Scannel entitled "OS/2 Users Head for the Exits," which purported to show a decreasing level of interest in OS/2 among corporate users. Yet the pie chart accompanying this article showed more companies increasing their investment in OS/2 than those decreasing it. And then there was the matter of a product comparison that needed a little tinkering.... but I'm sure it is obvious by now that plenty of "tinkering" has gone on at InfoWorld for several years now.

With magazines like InfoWorld blocking the channels of information between software vendors and decisionmakers, is it any wonder that IBM refuses to pander to their whims? How can IBM or any other vendor of quality products expect to overcome such a massive cultural and intellectual bias against innovation, against excellence, against the laws of engineering? Attempting to sell superior products to InfoWorld's readership is like trying to wake the dead.


[ Top | Home Page | Faq ]

Copyright © 1999, Tom Nadeau.
All Rights Reserved.