...making Linux just a little more fun!

<-- prev | next -->

Windows Defectors: Why Linux Is Worth Migrating To, Sometimes

By James Roberts

[James has agreed to coordinate the "Windows Defectors" series long requested in the Mailbag. This series is for those new to Linux who come from a Windows background. In spite of the series title, it's not just for those who have stopped using Windows completely. People who still like Windows, use Windows grudgingly, or deal with Linux-Windows integration in LANs will also find the information useful. Unlike LG's other series, this one will be written by multiple authors, since there is such an interest on writing on this subject. The next article in this issue (by Tom Brown) is also a Window Defectors article.

We're considering changing the series title to something less negative (e.g., "Linux For Your Mom", "Leaning Towards Linux", "Windows Immigrants Help Box", "c:\>, Drive-In for Windows Users", "At Liberty Under the Hood", "Linux Incognito" -- all suggested by Petar Marinov), but we're leaving it as "Windows Defectors" initially because that's what it has been known in the Mailbag. -Sluggo.]

"Those who ignore history are doomed to repeat it." -George Santayana

Have you heard that quote before? I expect so - and that's part of the point of the quote. This is the first article in a series aimed at the many people who are interested in Linux, but currently use Windows. It aims to ask, and perhaps answer, questions such as:

What's the context?

All such discussions have a context, and the context I intend here is that of the Small and Medium Enterprise, or SME. I'm not addressing the issues for the single desktop, or for the large multinational corporation. In fact, for both these extremes the issues are arguably simpler: at one end it's just personal choice of platform, at the other the resources are available to implement customised solutions as needed. But it is the SME area that has in the current, somewhat-recessive economic circumstances, consistently been identified by the IT industry as the last largely untapped potential growth area. I'm referring here to small installations of between two and a hundred desktops and servers. This SME area is price-sensitive, hasn't upgraded for years, has no IT management: it's where Linux can make a difference, and it's where I reckon it's hardest for Linux to enter - for reasons that I will describe later. If the problems can be solved here, then Linux will become 'sticky' at this level of the 'ordinary' user.

In this first article I'd like to set the scene, and this is why I started with the Santayana quote. I've noticed that people with a UNIX background often have had little contact with those who have been weaned on DOS - and vice versa. The target business that I'd like us to keep in mind is a small office - say eight or ten users altogether - and if this small office has a history in computing, it will generally have been in DOS and Windows.

So let's review where this typical office is now, and how it got here. It's a long and somewhat sad story, but in the end, you'll maybe see why, once, Windows NT was better for our clients than XENIX/SUN, and why now Linux may be better than NT.

Anti-flame note

Before we start, let me state that I am convinced that GNU/Linux is a masterly piece of work on the whole, and an astonishing achievement. Since it's made by humans however, it's not perfect, and inevitably it's the imperfections that I'll have to emphasise. The alternative options are also imperfect, that's why we're looking at Linux. But in some areas the alternatives may have strengths that GNU/Linux lacks. Please take the comments in the appropriate light. I also have to say that all trademarks are acknowledged.

Blast from Past

I'd like us to think back ten years or so - at which time (by mere coincidence) I'd just joined a small UK SME VAR as hardware tech. Within about ten days of starting I'd been bumped up into a systems and end-user support role as well. The VAR was then primarily a Novell reseller and customiser, and I quickly got to know the advantages and disadvantages of Novell 3.1. At that time most of the clients were solicitors (lawyers), and their systems typically were running a Novell server on a i386 or i486 with SCSI drives, with up to twenty-five diskless net-booted i386 workstations, running WordPerfect.

Overall the server/diskless workstation was a very good fit for the clients' needs; it was reliable, fast (the 10 Mbps network was faster than contemporary IDE drives) and easy to maintain (all the configuration files were in one place).

It was also abstruse, non-intuitive (indeed, counter-intuitive at times) and completely non-user-friendly. Any minor configuration error was often fatal. But it did work, and did the job well.

Enter the Band-wagon

Then Windows for Workgroups came along. We could put a network of twenty-five peer-to-peer workstations into a site for much less than we could do a Novell installation. Not only that, all the hardware was readily available and used standard parts (the diskless workstations needed specific types of PSU and weren't easily upgradeable). Moreover, with the Windows system, the clients had the illusion of a familiar system - most of the decision-makers who wanted to introduce networked systems already had Windows boxes at home. Also the clients could install their own extra software packages without invoking our aid (and paying for it).

Pretty soon we were installing a lot more Windows for Workgroups than Novell. Of course, it didn't take long for the downside to be apparent - no central installation of applications, no central control of rights, no user controls whatsoever, big problems with workstation backup, insecure peer-to-peer networking. So our recommended system included a Novell server, which solved some (but not all) of these problems.

Then Windows NT came out - and it was cheaper than Novell, both for the server and for the client access licences. And for the client, again it gave the illusion of familiarity - it looked like Windows. And for me, it came from the architect of Digital VMS - it was written by someone who did know what they were doing.

Meanwhile, Novell had changed their OS completely to version 4, which no one I knew found easy to use. They had introduced large-scale management tools, which merely muddied the waters for our typical installs of 5 to 10 clients, and it cost more than version 3.x.

Pretty soon we were selling networks made up of Windows for Workgroups workstations with a Windows NT 3.51 server, for a much lower overall cost than the previous Novell-plus-diskless workstations solution, and with pretty reasonable reliability and performance. Of course, building all the workstations took time, so I developed a custom cloning system that could build ten identical workstations in about fifteen minutes each.

Nixing XENIX

By then (and we are getting to 1995 or so), we had become pretty well-known and had a good reputation for looking after clients, so more than one solicitor came to us to replace their 286/386 XENIX-based system, which was by then getting pretty long in the tooth (like me now!). I never installed any RS232 terminals, but I removed a lot. They'd done a good job for their owners, but the replacement - with Windows 95 on the desktop and Windows NT 3.51 on the server - was like jumping forward a whole epoch, compared with the 9600 baud green-screen serial terminals in previous use. Was Windows actually as reliable? Well, yes, given the age of the systems being replaced, it was in fact more reliable.

So things went on, and as they went on the problems of this Windows computing paradigm - which had always been apparent in contrast to the Novell/diskless paradigm - became more and more cogent.

Windows 95 was effective - it's a masterpiece of forward and backward compatibility in my view - but it was not stable. It crashed easily. It was not secure. It was big, and the interactions within the systems were unpredictable. Add third-party software (and a different mix of such software on every box) and problems would arise that no-one reliably could solve. We got used to reformatting and reinstalling.

Windows 98 was better, and more reliable, but much bigger again, for no very good reason. It still crashed unpredictably, and took even longer to fault-find or reinstall. Windows NT server was pretty darn stable (after a few service packs), and so we all looked forward to Windows NT 4 coming in 1996, with the back-end stability of NT and the new interface of 98.

I adopted NT4 workstation as my personal desktop as soon as it was released, but I had to dual-boot with 98 because NT4 did not run games well - and I like games. However, NT workstation was too expensive to sell into our clients in general, so our standard install was Windows 98 on the desktop and Windows NT 4 server on the back end, and this worked pretty well.

Setting Sun

In 1996/7 we sold this system into various sites that were running Sun Solaris/SPARC. It was interesting how this came about. Why ever would someone using Sun SPARC workstations want Windows? Well, for freedom, odd as this may seem given the current context.

At the particular sites in question, 'the Sun' was under the exclusive control of 'The System Administrator' - who would install only what they wanted to install, after triplicate written requests, passed back up to company HQ and triple signed off by company IT. Maybe. Meanwhile the workers and researchers, who just wanted to get things done, were bringing in their own Windows machines and running SPSS or MathCAD or some custom sample analysis software on them just to get the work done, and meanwhile the new spectrometers in the lab came with Windows-based software that had to be copied BY HAND to the UNIX system! Yes, Windows and Windows networking at that time actually represented more 'Freedom' for these people, and eventually the lab Sun systems were retired and sold off.

Do I feel shame? No! I remember two issues. One was concerning a video lead for a monitor. The video lead - which if I remember correctly was standard apart from the blanking off of one pin - was $150, about, from Sun. I could source the same lead (apart from the blanked hole) for $7. And the price of a network card? Don't ask. Meanwhile, it took twenty minutes to link the lab pc talking to the machinery into the network - something that had been on request for a year. Any company that exploits a monopoly in this way develops users who detest it. Yes, any company.

Back in the future

So we get up to 1999 and Windows 2000 Pro. Windows 2000 is as good as Windows has got. It was pretty solid, pretty easy to use, fast, adequately tuned, and a good desktop system. But Windows 2000 Pro had a deliberately broken networking model, presumably to force a contrast with Windows 2000 server. Windows 2000 Workstation would only handle ten clients; this was reminiscent of the Intel tricks with the 486 series. The Outlook mail/groupware client was also broken, in that core functionality had been moved to Exchange server, and as far as I could see for one reason only: to make clients buy Exchange server. I didn't like this. But Windows 2K was still a good bet for our clients, although it was getting really overblown in size and slower and slower every issue.

The Plague years

Then the viruses started to really hit.

Then Windows XP came out.

We don't sell XP, if we can help it.

We have a few problems with it. We don't like the licence (it's a potential problem for most solicitors/lawyers and anyone else with a duty of confidentiality), we don't like the desktop (it's an insult), we don't like the vulnerabilities, or the slow performance, or the sheer undocumented mass of it all. But that's not the real problem. The real problem we have is that soon we won't be able to get Windows 2000 any more.

There's a problem in another area too. Many of our clients are running Windows NT 4 server. It's very solid. It works. It does the job. But there will soon be no more security updates or support for this version, and we will have to migrate clients to something else. Windows 2000 server, 2000 AS or especially Windows Server 2003 are a ridiculous overkill for an office with six client machines, as well as being much more expensive and needing huge hardware resources to run. So what else can we install to replace NT4?

Enter the Penguin

Now, back in 1993 I had ran into Slackware, presented as a cover disk with Personal Computer World UK. I installed it and played with it, but it didn't do anything I needed. Nonetheless, I kept an eye on GNU/Linux and installed most builds of Slack, then Red Hat. We bought in a pre-configured GNU/Linux Slackware box in about 1996 to run as our mail server, initially on dial-up and then ISDN, and over 5 years it never crashed. The PSU failed one day, and I managed to migrate the system to a different box and a bigger hard drive and get it all working again (I never was sure how). Many people say that Windows is unstable. With the NT series and all relevant patches on supported hardware - I just don't think that's true. I've had Win NT 4 boxes stay up for years. But that GNU/Linux could so easily match or exceed this performance impressed me.

Moreover, GNU/Linux has over the last few years transformed its graphical desktop capacity. Not only are there lots of interestingly geeky X-desktops available, but two mainstream competitors for Windows 2k/XP, with generally enhanced capabilities.

Two years ago, with the launch of XP and the developments in Samba, I decided to seriously evaluate the potential of Linux for replacing Windows NT 4 on the backend with a GNU/Linux box running Samba. I installed every mainstream distro I could get, on every box I could test it on. I played with it, broke it, fixed it (sometimes) and did my sums. Then we put some test boxes into clients, mostly as mail servers/file servers with AV integrated, and they have on the whole been working flawlessly ever since.

But all is not sweetness and light. There are some problems to overcome in retro-fitting GNU/Linux into a Windows network, and now I've given the history I've just laid out, I hope you'll see them a little differently.

Next month I'll start discussing what sorts of problems we've encountered and how we've tried to work with them.


[BIO] James is the coordinator of LG's "Windows Defectors" series.

Copyright © 2003, James Roberts. Released under the Open Publication license unless otherwise noted in the body of the article. Linux Gazette is not produced, sponsored, or endorsed by its prior host, SSC, Inc.

Published in Issue 97 of Linux Gazette, December 2003

<-- prev | next -->