A response to Unix vs NT

By Laurent Caillette, french software engineer.

First of all, I'd like to express my admiration towards the writers of this high-level article, a sound of truth among a sea of lies and unconsistency. Some may think this article is just bill-bashing. It is not. Unix is not a perfect world, but these revelations about the lack of quality of NT seem too enormous to be true.

In fact I think they are. I've not always been allergic to Microsoft products. It even happened to me to say : this one should be great (while reading the paper). I've been deceived several times (apart of Project 98 and FrontPage 98) and NT is no exception.

These are some meditations issued of my short professional living.


I really loved this part of the article of John Kirch :

Why Windows NT Server 4.0 continues to exist in the enterprise would be a topic appropriate for an investigative report in the field of psychology or marketing, not an article on information technology. .

That's it.

Windows is a ultimate form of demagogy. Many users love it because it's immediate to use, and behaves not unlikely a toy at work. I'm often surprised by the reaction of users who face the loss of one-week-long labour, due to a system crash, or a virus attack. They've got half a smile on their face, like the survivors of a minor catastrophe. They recognize themselves in a kind of drama, seen on TV. Since War Games (the movie) and the installation of Windows on their compu ter, IT will never be far of a play-place for them.

Autistic IT people

Some responsible of this lack of professionalism are IT people themselves. They don't know how to refuse to do impossible things. If running mission-critical operations on NT was impossible, it wouldn't be. That's all. In fact there's no real impossibility. The bug is it costs (really) more.

IT people who want to get NT out should change their habits and integrate a fearful weapon in their arsenal : the cost study.

It's a part of our culture to refuse to be time-controlled. In fact, there are some good reasons. It's always difficult to announce to your boss that your job took twice more than planned, for, most of time, ridiculous reasons. The most beautiful time schedule is close to be a white one.

Forget. It should change. Those who want NT out have to follow their activity, hour by hour, especially around troubleshooting, and make s tatistics. A real boss is very money-aware, especially for money lost. A bill will be more credible than your speech.

You don't like the idea ? It's not in your habits ? It seems dangerous ?

You prefer to stay with NT ?

Those who work at Redmond

Nowadays, everyone's OK to say that the authors of MS-DOS were almost a kind of hackers, nothing to do with the high-level architects of the XXIth century mission-critical operating system. It would be instructive to learn a bit more about their time-schedule for last ten years.

It's obvious they still worked inside Microsoft. They didn't spend their time in a university to learn what a well-written API looks like, nor a security system in a real world. In fact, they're those behind the main specifications of NT.

The capacity of changing of Microsoft is always overestimated. In fact, the engineer still belong a microcosm, werhe OS is believed to be something that fits for a lonely hacker, who's got very much time to spend with it. They learned to make compromises, but it's impossible to change one's mind at all. Don't believe press announcements. Just believe your eyes.

The authors of VMS are said to have worked on NT. Not at the top level of decision. I don't know VMS at all but it could'nt have been shipped with the following options :

  • Pending shortcuts when target is destroyed / moved
  • Incredible ease to move a system directory with a simple click
  • ...And many inconsistencies listed in the article referred to

As a matter of fact, VMS's authors brought their help, but I don't think they were at the top decision level. NT 4.0 looks too much like 95, and It's almost Microsoft's interest to make us believe that NT's true authors are VMS's ones. Because it's not really true.

Do you know many people who would pay thousands of do llars for the DOS revenge... by the same authors ?

Two ways of thinking an OS

I say : a new OS architect should be convicted to own the right solution to make people working more efficiently. If this guy is clever enough, if he's given enough money, and if nothing goes wrong, then the result should be great.

Microsoft's efforts are ruled by other views. They may ask themselves something like : "what is the minimal features set which makes our product a killer on at buying time ?

It's another world. NT end-user is not a IT user, but... a NT reseller. Such an ugly thought !


NT makes companies loosing money. But who saw those bucks burning ? In fact, they're coming in other's pockets, who are :

  • IT people who can get sure to insure their seat
  • Teaching compagnies
  • Software engineering compagnies, especially those which sell software to patch NT's lacks
  • Press publishers
  • OEM's who make benefit selling NT

NT's ability to generate entropy should be understood as a major reason of its success. All those who earn money should hope that it will last, and do the best they can to contribute to NT success.

NT's real cost is obfuscated by each one's effort. IT managers affirm that NT is stable and not error prone, and makes deployment easy. Then all workload estimation will get wrong, because they should reflect a manager's will instead of reality. Software teams will solve NT problems with more hours at the office.

And those who spend extra hours, digging into the ugly MSDN, will avoid to tell it because :

  • They're afraid to look unaware
  • They think that the choice of Microsoft products will never be reviewed
  • Learning a bit with this tool has been just a pain, they're no t ok to do it with another tool
  • It has been such a pain, that their little knowledge costs gold

So I can't imagine conspiracy members supporting a migration towards any kind of Unix. Especially if it is a free one !

A true story of mine

John Kirch's article concedes that NT may be used on workstations. Maybe sometimes, but sometimes not.

I develop on NT40 server, with a 64Mb Pentium II. It crawls. I've to reboot twice a day, when I terminate programs abruptly. Our app has been deployed over 200 sites. The client side runs over NTWS, it is a high-end cash register.

Deployment cost

We had to set a remote access system. Facing the lack of Telnet, we installed a PC Anywhere on each client. PC duplicates GDI commands on the slave station. The user should stop working during the intervention, and the screen refresh speed is low enough to drive any hotliner crazy.

This system doesn't fit, but costs much. Let's sum up : NT licence, PC Anywhere, antivirus. An install procedure needs somebody close to the PC, just to click on the 'OK' button at the right time. The master disk may look like the solution. But the master disk works only if all hardware configurations are exactly the same. With different customers and different needs, evolution of hardware, drivers updates, configuration management should become an interesting job soon.

Development cost

One could say, at first glance, that NT was a suitable platform for development. Delphi 3 and Interbase were reliable and gave us a boost we couldn't expect from... VC++ (for instance). In fact, I don't know any framework nor debugger like Delphi's ones in Unix world.

Nevertheless, NT was a heavy burden for developers, and I doubt it saved us so much time.

  • Finding how to kill a process, knowing only its module name : 3 days
  • Emulating a keypress into an aggregated ActiveX control : 5 days
  • Fixing the bugs of OPOS (Ole Point Of Sale, a standard API for retail peripherals) : 1 month

The developers involved had at least a two-years long experience on Delphi. The fact is, Windows API is a vast garbage where you can't find what you need. And MSDN is not better.

So Linux ?

Ignorance was the real author of NT choice. Ignorance of other products, and hidden costs.

Was Linux the solution for us ? I think so. Development time would have tripled (the cost of replacing Delphi), but it's not the worst for a compagny making most of its benefits on support. I'd like this experience to enlight people having to build app for dedicated hardware.

Not for everybody...

We shouldn't expect a tsunami whipping out NT for next two years. IT managers suffer of too much inertia, and all tho se who make benefit whith NT will do their best to keep people on track.

I'd like to explain briefly my very personal regrets.

...And not too soon

But I should admit that Linux phenomenon should be preserved, by a long-term assimilation of incoming users. Newsgroups, in John Kirch's article, is presented as free hot-line. I don't agree. Free resources have to be protected, because Linux developer's good will may decrease as they discover their job is borrowed by companies selling products for good money.

Luckily, the powerful motion around Linux and free software is still intact. Announcements from Sybase, Oracle, SAP are more anoblishment for Linux supporters than real threat at this time.

I hope seriously to meet soon a balanced world where :

  • Managers would agree with the choice of Linux
  • Linux would continue to grow, supported by thousands of developers making it better -- for pleasure
  • Companies under NT would be aware of the consequences of their choice
  • There would be enough jobs for NT programmers feeling the need to go to the other side

Incomers need to be progressively integrated. I'm one of of them. I think that the main threat is that business may ployder good and generous ideas.

Linux's lack of legitimity is its best defense at this time. Let's take enough time, to organize the economical model which would preserve the original spirit of free software.

[back to the beginning]