A couple of years ago, I wrote OSWars 98 which covered the major Oses than out. Since then, I have regularly received requests to update it to reflect the changes in the OS market. Unfortunately, this has become increasingly difficult as many of the OSes have increased their rate of improvement making it hard to give a fair comparison of the Oses. Afterall, there’s always a new version right around the corner that will change everything.. Well, enough’s enough and I’m just going to draw the line with what’s out now.
One other thing that has become a much more important distinction is that I must stress that I talk about desktop Oses. I am not going to get into comparing these Oses in terms of server abilities. This document refers purely at their ability to provide a personal computer user or work station user a way to get the most common jobs people use PCs for.
Oses to compare are:
that should be included but can’t due to time or availability:
MacOS X with Aqua
Windows 2002 (whistler)
I’ll talk about the above in passing, or more accurately in drive-by mode…
I still don’t care for Windows 9x. While there are those that claim it works like a champ, it is still remarkably unreliable. Let me put it this way, MS Word 2000 crashed on me for no apparent reason while writing this the first time (luckily I only lost 10 minutes). I’d be in Windows 2000 right now but the laptop I’m writing this on only has 64MB of ram.
If your computer can’t handle Windows 2000, Windows 98SE is probably the best choice for desktop users who work with a variety of applications. As an OS, Windows 98 is pretty mediocre. It crashes, its registry tends to get hosed, it has a resource limitation causing apps to become unreliable when resources get low, and it’s just plain flaky. I don’t know how many times a problem will just disappear if I happened to be running Windows 2000 instead of Windows 98. Still, it’s the cheap, mass market OS of the masses and therefore has complete hardware and software compatibility. Basically any mainstream program worth having will run on Windows 98 and any mass market hardware product will run on it. But you take away that hardware and software support and it’s a pretty rotten OS. I am not quite sure what difference Windows 98SE (second edition) has over plain Windows 98 other than bug fixes.
Figure 1 Windows 98 with Object Desktop's WindowBlinds (www.objectdesktop.com)
I’m not a big fan of Windows in general due to its one-size-fits all design but I have to hand it to the imperial developers at Microsoft that Windows 2000 is pretty magnificent. If Microsoft didn’t tend to behave so evilly so often they would get even more praise on their achievement with this remarkable OS.
Not only is Windows 2000 more stable than Windows NT 4.0 by a long shot, it offers the game compatibility, plug and play features, and much of the hardware support that can be found in Windows 98. At our offices, every hardware device worked great with Windows 2000 except for the HP scan-jet we had (HP didn’t have any plans to support their scanner under Windows 2000 despite it being for sale presently).
Despite some claims to the contrary, basically any new game will work great on Windows 2000.
The new fragment resistant NTFS 5, like its NT 4.0 predecessor supports compressable drivers and folders which can save many megabytes of disk space. It also provides a new partitioning system that has “Dynamic disks” with “Basic” partitions. The practical use of this is unclear to me thus far. But I do know that a pure NTFS system does not suffer from broken short cuts. Move the original file and a short cut is automatically updated. And thanks to the inclusion of “hard links”, this system is much more robust than the fragile “Shadow” system in OS/2. Moreover, Microsoft seems to want to push what I call “Super Extended Attributes” that one day will be used for creating apps with much better ways to organize and sort data.
Figure 2 Windows 2000 with Object Desktop
Still, not everything is perfect. Like all versions of Windows, it has the one size fits all problem. Microsoft has already decided how your desktop should look and behave. Want to get rid of “My Computer” and put it in its own folder? Tough, you’ll need to monkey with the registry or get a utility like OD to do it. Windows 2000 gives the user remarkably little flexibility out of the box. While this creates a great market for third parties like Stardock, it can still be annoying when you bump into something that hasn’t bee solved by third parties.
Stability-wise, Windows 2000 is great. It is rock solid. My desktop goes weeks at a time without reboot. A big over NT and Win98 which seem to demand that you reboot for virtually any minor reason. Windows 2000 almost never needs a reboot and even when a program says you need to reboot, in reality a logoff will accomplish the same thing.
One unfair charge that gets leveled against Windows 2000 is that it’s a “resource hog” because it needs 128MB of ram to run well. This is the same kind of strawman argument people used to throw at OS/2 because OS/2 needed 8 megs of ram whereas you could multitask on an Amiga with only 512K. In 1993, a 4MB SIMM cost between $120 and $150 depending on the day. OS/2 2.1 needed 8 MB of ram to run decently. That meant that the ram requirement cost at least $240. Today, however, you can buy 128MB DIMM for $120. So for the same cost it used to cost to get OS/2 to marginal performance, you can get 256MB of ram making Windows 2000 have about a couple hundred megs of RAM free to use on apps.
So is Windows 2000 “bloated” Not for what it does. And from an expense standpoint, which is what matters in the real world, Windows 2000 is only half as “bloated” as OS/2 2.x/3.0 was during its time.
Linux is the hardest of the OSes to fairly evaluate. That’s because where exactly do you draw the line on in terms of features when so much of it is free? Do you count just what comes on the CD or do you count all the free open source software you can get?
Also, bear in mind that this OS comparison is about desktop operating systems. I.e. the type of OS a corporation would use for their users down to your neighbor with their $1200 Dell PC. If you’re a Linux advocate, please keep this in mind. Linux is a great server OS but we’re not talking about server OSes here (OS/2 is a great server OS too btw).
Let me first start this portion by advocating Linux in a way that I don’t see it advocated. In my mind, Linux could really take off if the commercial companies involved on it began showing off the power of seamless distributed computing. This is one of the big kicks I’ve been on for the past couple of years. With more and more people on 100Mbit and higher networks, it seems to me that I should only rarely have to wait for things to be finished. Compilers, rendering packages, graphics applications, and anything else whose tasks could be divided up into threads of execution could be seamlessly split over a high speed network. Applications could easily be launched anywhere on the network and simply display on your machine and so forth. Linux can do much of this today with a combination of custom applications and X Windows. But I don’t think this is exploited enough and the open source community, in my view hasn’t done a great deal to make setting things up to make Linux a trivially easy distributed computing network. Before someone sends me a patronizing email (as Linux “advocates” tend to do) on how “easy” this is, we are talking “easy” in the sense of anyone even vaguely familiar with computers being able to set it up. Not “easy” as in someone who’s used to configuring the script to get their modem to work to do it.
Figure 3 Linux with KDE (www.kde.org)
So that I could ensure fairness in this topic, I went over to Nicholas Petreley’s (editor at www.linuxworld.com) house and he was kind enough to show what Linux could do as a desktop that can benefit “power users” who are largely the kinds of people who would be reading this article:
The application support on Linux is pretty good. For getting most things done, you can do it on Linux. Unfortunately, in my view the applications are generally inferior to what is available on Windows (and OS/2 even). For instance, while Gimp is a very impressive free application, it’s no where near being up to par with Photoshop or Corel Photopaint. Star Office is a good suite but I don’t see it as being any better than MS Office which, we have to face, is what most people are using and it doesn’t read in Office documents very well. Word Perfect is kind of available but it mostly uses WINE to do its magic and it's hard to tell whether Corel is supporting Linux as its latest fad or whether it's really serious about it.
The one thing that particularly impressed me was how quickly I can switch between different user sessions. I could be logged on with Root but be primarily in my Bwardell account (i.e. I don’t use Root as a user but I might have it logged in to do some maintenance and such).
Driver support is pretty good, better than OS/2 and Be, a bit worse than Windows 2000 and much worse than Windows 98. Odds are your hardware will work with Linux but odds are that any special features of your hardware won’t be used. One could say that Linux driver and app support isn’t as good as Windows have nothing to do with the OS itself. But the OS is merely a means to an end. It doesn’t matter how good the OS is – without apps it’s useless. I would argue that in 1994, OS/2 was vastly superior to anything else out there from a desktop OS perspective but it didn’t get the application support.
Linux, unlike the other OSes, is not limited to purely x86 hardware. I don’t really consider that a big deal from a desktop perspective but from a server perspective that can be a huge deal.
But now comes the bad news. Linux doesn’t have anywhere close to some pretty basic features that are required for it to succeed as a mainstream desktop OS. Open source and distributed development has been Linux’s biggest advantage since certain predatory business tactics won’t work on it as a result. But it is also now limiting it. Here are a few basic examples:
These complaints aren’t to say that people shouldn’t use Linux as a power user desktop OS. It just means that Linux would be a pretty poor choice as a general user desktop OS unless you need one of Linux's specific strengths. However, let me point out that I think with some work on the polish issue, Linux could do the job really well for a casual user who primarily wants to just do web browsing. They need to get anti-aliased font support in, however. First thing I notice on a Linux machine with Netscape is how ugly the text is (at 1600x1200 it's not so bad but not too many people have a 21" monitor to enjoy that).
While third party software support isn’t necessarily the fault of the OS, it’s immaterial. OS/2 and BeOS, for instance, are at least as good if not better desktop OSes than Linux currently is in many ways but lack the third party support. Linux has better third party support than OS/2 and BeOS and has the momentum behind it so one can forgive Linux’s rough edges.
But Windows 98/2000 lacks the rough edges and has the third party support. This means that when it comes to using Linux as a desktop OS I can only see a few reasons why one would choose it over Windows:
1) If you hate Microsoft and want to have a Microsoft free computer, Linux is a good choice.
2) You are on a high speed network and want to do seamless remote control of applications and distributed computing.
3) You need to have “real time” OS in which you have utter control over the priority of the processes to ensure rock solid stability.
4) You really really hate to pay for any software and think it is somehow wrong to charge money for intellectual property.
There are other reasons but these would be the big 4 in my mind. I think one of the problems also hurting Linux is that the community there seems to dislike the idea of commercialization. There are a fair number of folks who incorrectly think that the Internet we use was made possible by some sort of free intellectual community or something when in fact it began as a government project and took off when corporations were allowed to get involved. Anyone who reads Slashdot regularly as I do knows that the vocal portion of the Linux community has utter disdain for commercial software companies in general.
Figure 4 Linux with Gnome (www.gnome.org)
Obviously, I’m biased as an employee for a commercial software company but I don’t think Linux can really become a mainstream desktop OS without commercial companies coming in. But the smaller companies out there and even the shareware authors I’ve talked to are wary of Linux because they can’t find a way to survive by giving away their software for free. This in turn has scared off a lot of small developers who don’t like writing for Windows but feel they have no choice away.
In conclusion, Linux is a great server OS but not the best desktop OS now. It will be interesting to check in a year or two and see how they’ve progressed in the next couple of years. I seriously think that the next time I do this report, the gap between Linux and Windows will be so small that it'll come down to just application taste.
The current version of OS/2 Warp for desktop users is still 4.0. Now at Fixpack 13, this aging OS is still used on numerous corporate systems and by a handful of enthusiasts. But it’s a far cry from its glory days of 1994 when Team OS/2 worked Comdex demonstrating to an eager public the power of preemptive multitasking, robustness of the OS, and its unique object-oriented shell.
Unfortunately, by any reasonable standard OS/2 is technologically inferior to the other OSes in all but a few areas. While OS/2 has always had a problem getting third party software support, now it is driver support that is becoming an issue (except in the video department where Scitech's heroic efforts have actually put OS/2 in really good shape).
However, USB, DVD, SMP, and sound driver support either don’t exist or exist only in the most nominal way. By nominal let me be more specific: OS/2 with the latest Fixpacks can support certain USB devices like floppy drives and some printers. But most people use USB for scanners, digital cameras (OS/2 does allegedly support a few of these), video cameras, etc. Hooking your printer up to your OS/2 box with a USB cable is swell but not particularly interesting unless you’re using that parallel port for something else already. And by DVD “support” OS/2 will eventually support the ability to read DVD drives like one reads a CD data drive. But that’s a far cry from playing DVD movies and such.
Figure 5 OS/2 Warp 4 with Object Desktop
What’s killed OS/2 the most as a desktop OS is the lack of an updated client. Installing OS/2 on modern machine requires a great deal of driver hunting just to get it to install. In an age where BeOS can be installed in a few minutes and that one can simply insert a Windows 2000 CD in a drive and reboot the machine and have it install, OS/2’s 4 year old client CD makes it an unreasonable choice for users new to OS/2. This makes OS/2 strictly a legacy platform – i.e. only those who got into OS/2 in 1996 and earlier are likely to be using OS/2 today. I suspect that the next time I do an OS comparison, it won’t make sense to even include OS/2 no matter how much sentimental love I may have for this once state of the art OS. There is some modest hope, however. Serenity Systems is hoping to create a Warp 4 based consolidated client CD. I am not sure how much this will help (at this point, OS/2 needs a major update on the desktop side if it's going to have any chance at growing) but for those people still on OS/2, this can only be considered good news.
Still, OS/2 still offers some unique features for the desktop user. Its object-oriented shell (The Workplace shell) which is ironically not supported by IBM anymore (nor its underlying technology, SOM) is still first rate even if poorly implemented. Its objected oriented shell allows for productivity minded users to greatly increase their productivity by organizing their work in a logical fashion.
Sadly, OS/2 is a dying animal on all fronts and arguably already dead as a desktop OS. Without a way to attract new users, there is little incentive for software developers to create software for it. And IBM’s recently released OS/2 strategy makes clear, for the first time, that IBM wants its customers to move off of OS/2. While IBM continues to make a significant amount of revenue selling services to its OS/2 corporate accounts, it’s become enough of a distraction for that they’ve finally outright told these customers to move their applications to a platform neutral platform like JAVA and then move to another OS (such as Linux).
If OS/2 is mentioned in OS Wars 2002 as its own section, I’ll be very surprised.
With BeOS 5, Be, Inc. made a free version available to the public. Too little too late? Only time will tell. BeOS has a great deal of impressive technology and a relatively clean slate to draw on. Its proponents will tell you how fast it loads, how little memory it consumes, and how reliable it is. All true but then again, OS/2 1.3 was all of these things too back in 1990. An OS with little software on it has little opportunity to be pushed to the limits.
Figure 6 BeOS in action
That is the obstacle Be has to overcome still. What are people supposed to do with it? What does Be over say the Mac or Windows 2000? Certainly not in applications. One of my complaints is that BeOS doesn’t seemed to have been designed with distributed computing in mind from the beginning. This is one area that Be could have found a niche I believe – a true media OS could have had its killer app as a powerful rendering package that could be clustered together painlessly. Linux was used this way on Titanic but it was probably a considerable pain to setup compared to what Be could have done. With 100megabit LANs becoming common and even higher bandwidth networks becoming readily available, an OS designed to distribute a task throughout the network on the fly seamlessly could have made a potent niche to grow from.
That aside, I like BeOS, I just can’t find that much to do with it or enough compelling things to do on it over Windows, or OS/2 or Linux to make me switch. But I can see why someone would run it. If you hate Microsoft, think OS/2 is dead and don’t want to putz around with Linux, BeOS is a good alternative.
In January, Steve Jobs demoed the new MacOS X with its Aqua interface. It created some serious momentum for Apple’s next generation OS. The impessive look of Aqua was quickly copied by Linux and WindowBlinds “Skinners” bringing the look to the PC within hours of the first screenshots being posted.
But MacOS X is a lot more than a pretty user interface. It’s a totally new look at how a computer should be used. In my opinion, Steve Jobs’ goal for the Mac is to make it into a true appliance in the same way that a VCR or a refrigerator is used. Easy to set up, easy to maintain, and easy to work with. I think they’re onto something if they can keep the downsides to a minimum.
Mac OS 9 is a fairly primitive OS technologically. With pre-defined memory spaces, no crash protection, and “cooperative” multitasking, Mac OS 9 is akin to Windows 3.1 technologically. Its UI has always made up for its deficiencies but as Windows and Linux have moved forward with things like Gnome on Linux and Object Desktop on Windows, the UI of the Macintosh isn’t enough to save it alone. That’s where MacOS X comes in. Basically NeXTStep with a new UI, it provides preemptive multitasking, memory protection, and the other benefits of a modern OS. How the Aqua user interface works in practice remains to be seen.
I really think Microsoft should have just skipped this. Do we really need yet another upgrade to what amounts to DOS with an integrated shell? Windows 98, Windows 98SE and now this? People lived with Windows 3.1 for 4 years. I think they could have handled Windows 98 for a couple more years until people had migrated to Windows 2000 or Windows 2002 (Whistler).
No sooner does Windows 2000 show up that we’re told about the next great thing. Early reports say that Whistler is very nice. Improving legacy compatibility with limited distributed computing integrated (basic level of MS Terminal Server included), Whistler looks impressive.
My problem with Whistler is with the Microsoft way of doing things. They are so reactionary that they rarely if ever actually innovate. Microsoft Media Player 7 has skinning now. That’s great but did they really need to do that? Winamp and Sonique filled this niche nicely but it seems like Microsoft will look at the top 20 downloads and begin making a version of whatever software that might be. Zip, MP3 players, instant messaging. While some might consider this a good thing, it hurts users in the long term by stifling innovation.
There are a ton of Linux variants now. I am not a fan of Linux on the desktop. So I went over to the house of one of the most knowledgeable people on Linux – Nicholas Petreley, editor of Linuxworld (www.linuxworld.com) and the author of “The Open Source” column at Infoworld.
He gave a pretty objective demonstrate of the strengths and limitations of Linux on the desktop (I believe he was using Caldera Linux but I could be wrong). I’ll talk more about that in draft 2.
I would like to hear from any Linux advocates on why they would advocate Linux as a desktop OS over the alternatives using reasons beyond “Because I hate IBM and Microsoft and Apple”.
The state of the art in desktop operating systems has really improved over the past 2 years. Plug and play now actually works on Win98 SE and Windows 2000. USB is here even if it is limited to only 12Mbit right now.
Most of them have improved themselves when it comes to installing them (particularly Linux) and have improved in terms of ease of use (again Linux). I would have to say that of all of them, Windows 2000 is probably the best desktop OS overall. It has the power, stability, and ease of use that most people would need.
For casual users, I have to regretfully say that Windows 98SE (and soon Windows 98 ME) is probably the best choice despite its flakeyness. It’s amazing how many of “our bugs” in a given product (whether it be a game or utility) magically disappear when the user upgrades from Windows 98 to Windows 2000.
Linux has come the furthest as a desktop OS but it was so far behind that now it’s merely a distant third instead of being completely unreasonable as a desktop OS as it was in 1998. With some polish and desktop based standards I could imagine seeing it on machines in a typical office or home.
hasn’t changed since 1996 realistically and it’s shown.
What was once the world’s greatest desktop OS is now basically a
legacy OS with a few die hard users left.
BeOS is a good OS but just isn’t getting the software support it needs. It’s having a hard time finding its place. It doesn’t have that “killer app” or “killer feature” that would make people use it over Windows or Linux (or Macintosh).
of you who like to make your computers look and feel like other OSes actually
have a few choices:
On Windows you have Object Desktop with WindowBlinds:
If you already have WindowBlinds, you can download the OS Skin pack here.
On Linux you have Enlightenment and other skinnable WMs with skins (themes) at:
On Macintosh you have Kaleidoscope at:
On OS/2 you have Object Desktop at
On BeOS there is a keyboard code that will change your OS to look like other OSes (but I don’t know it off hand).
Hope you liked this report. Please feel free to repost it and use it as you wish. Just give credit when you do and if you wish to use this in a publication, please contact me at firstname.lastname@example.org. All I ask is a copy of the publication when it comes out if possible.
Brad Wardell is the Product Manager of Object Desktop at Stardock Corporation's ASP (Stardock.net). Wardell specializes in GUI design and user environments. Stardock's home page is http://www.stardock.com