February 2000 Column

[ Site Index] [ Linux Index] [ Feedback ]

Bubbling Trouble

There's been a lot of heat and very little light generated over the Microsoft anti-trust case recently. You're probably bored of it by now, and even more bored of reading prognostications about how it's all going to be settled -- whether Judge Jackson is going to order the company to be broken up or select some other remedy, whether Microsoft will appeal, whether they'll win the appeal, and whether it actually matters.

Well, this isn't the most important lawsuit Microsoft are embroiled in right now. The DOJ can't destroy Microsoft; all they can do is compel Microsoft to pay attention to their public image and not wilfully trample on competitors. A break-up may even make Microsoft's shareholders richer -- that's what happened to AT&T, after all.

On the other hand, there's a lawsuit in progress that has the potential to really damage Microsoft -- possibly to force them into a steep decline. And it's relevent to this column because the plaintiff bringing it is actually a Linux company -- Caldera Systems.

The Caldera lawsuit dates back to 1996, and it wasn't started by Caldera; it was started by Novell. Novell had just bought a company called Digital Research. DR was the outfit that sold CP/M -- the main personal computer business operating system of the 1970's. It was Gary Kildall of DR who famously was out flying when the blue-suits from IBM came to buy an operating system for their new PC. During the 1980's DR signally failed to compete successfully with Microsoft, whose MS-DOS started out as a thinly-disguised clone of CP/M-86. But around 1990, DR's flagship product, DR-DOS, provided a better DOS than MS-DOS; while MS-DOS was a stable cash cow, DR were agressively adding features and improving their offering. DR-DOS shipped on numerous PC's as the operating system of choice, but by the time Novell acquired DR in 1996 it was a spent force; Windows 95 was out and didn't need to run on top of DOS.

Or did it?

The lawsuit, which Caldera inherited from Novell when they acquired DR-DOS and CP/M -- Caldera's main investor is Ray Noorda, founder of Novell, and both companies are based in Utah -- alleges that Microsoft deliberately tried to kill DR-DOS. Windows 3.0 ran on top of DR-DOS without problems, but Windows 3.1 complained about bad device drivers. Moreover, Digital Research engineers were not allowed to participate in Windows 3.1 beta testing (although they'd been involved in the 3.0 trials). There were hints of darker skullduggery -- deliberate attempts by Microsoft to tie their GUI to their own version of DOS -- and around the end of 1998 and beginning of 1999 Caldera's lawyers managed to subpoena various internal Microsoft documents. There's a log of the proceedings on the web at www.drdos.com, and it makes fascinating reading.

Microsoft retaliated in February by filing for summary dismissal of the lawsuit -- asking the judge to throw it out as having no merit. Judge Benson declined to do so. Meanwhile, Microsoft attempted to have a lot of the subpoenaed documents withheld from the public record on grounds of confidentiality. They failed -- and what we are seeing is rather entertaining, in a gruesome sort of way.

"You never sent me a response on the question of what things an [application] would do that would make it run with MS-DOS and not run with DR-DOS," Bill Gates wrote in 1988. His earlier email? "I am not looking for something they cant get around. I am looking for something their current binary fails on." His reasoning? "First, we have to make sure Windows isn't easy to clone for both technical and legal reasons. ... DOS being fairly cloned has had a dramatic impact on our pricing for DOS. I wonder if we would have it around 30-40 % higher if it wasn't cloned. I bet we would!"

According to an internal Microsoft memo dated June 16, 1992, "Novell is after the desktop. As you know, they have acquired Digital Research and are now working hard to tightly integrate DR-DOS with NetWare. We should also assume they are working on a Windows clone and/or that they are working on a virtualised DOS environment which will run standard mode Windows as a client. This is perhaps our biggest threat. We must respond in a strong way by making Chicago a complete Windows operating system, from boot-up to shut-down. There will be no place or need on a Chicago machine for DR-DOS (or any DOS)." And in case this all sounds like business as usual, Judge Benson released the following unattributed snippet from a senior executive: "If you're going to kill someone there isn't much reason to get all worked up about it and angry--you just pull the trigger. ... We need to smile at Novell while we pull the trigger."

All of this is fairly slimy stuff. But there's more. Microsoft went after DR's customers, too. German PC manufacturer Vobis was lured away from DR by strong-arm tactics; Microsoft told them that they'd be charged $35 per copy of Windows, but only $26 if they bought Windows with MS-DOS for each machine they sold. Microsoft even went so far as to pay $50,000 for pre-paid copies of DR-DOS that Vobis had already purchased.

And then there's Windows 95. A interesting facet of this case is that apparently the attempt to render "Chicago" (Windows 95) inoperable on other versions of DOS wasn't entirely successful. You can boot Windows 95 on top of DR-DOS -- if you modify DR-DOS slightly to support an undocumented system call Windows issues at start-up (and at no other time). It looks as if the "DOS-less" Windows 95 (and, by extension, Windows 98) is anything but -- and Microsoft appear to have rigged it in this way specifically to kill a competitor.

Incidentally, the parallels with th DOJ anti-trust case are striking. The DOJ alleged that Microsoft spliced their web browser into their operating system in order to freeze a competitor (Netscape) out of the market. In this instance, Caldera are alleging that Microsoft spliced their GUI into their operating system in order to freeze out a competitor (Novell).

As Judge Jackson wrote in his findings of fact in the DOJ v Microsoft lawsuit: "The cumulative effect of the stratagems [ browser integration ] was to ensure that the easiest and most intuitive paths that users could take to the Web would lead to Internet Explorer, the gate controlled by Microsoft." In the desktop operating system struggle, Microsoft wanted to kill Novell in the server market. Their tool in doing this was Windows 95, much as their tool in attacking Netscape was Internet Explorer and Windows 98.

The next Caldera v Microsoft hearing is due on December 7th, to set a date for trial; Judge Dee Benson has a full book until January '2000, but the trial almost certainly will go ahead that year, in front of a jury in Utah. Caldera are asking for damages estimated to be on the order of four billion dollars, and if Microsoft are found guilty punitive damages of three times that may be awarded. Given that the DOJ trial findings of fact are a matter of court record, it's possible that those, too, will be dragged into this trial; in which case Microsoft are painted into a very tight corner.

Two possibilities arise if Microsoft lose this case. The first is that Caldera will suddenly mushroom in size or begin to make waves that splash right over the walls of the (currently rather small) Linux corporate world. (What does a Linux company do when enough money to buy all its competitors twice over falls into its lap?) The second is a flood of class action lawsuits against Microsoft. Suddenly Microsoft will be fair game for every ambulance chaser in town.

And as if this doesn't suggest Microsoft are in trouble, I haven't gotten onto the subject of the ongoing French investigation into allegations of criminal fraud and violation of EU trade law on bundling of products. Expect 2000 to be the year of lawsuits -- and the year when Microsoft has so much on its plate that attempts to target Linux with FUD will be severely restricted by the necessity of looking reasonably honest in court.

It's over.

In January 2000, Microsoft and Caldera agreed an out of court settlement, days before the case came to trial in front of a jury.

Details are covered by a non-disclosure agreement. Microsoft announced at the time that they'd be taking a one-time hit on dividends, from which the less astute observers calculated that they'd be paying Caldera roughly $160M in damages. However, Microsoft is cash-rich; they can in principle afford to pay ten times that without reducing the dividends they pay their shareholders. As Caldera were suing them for an amount on the order of four billion dollars, I personally think it is unlikely that they'd settle out of court for as little as $160M -- we'll probably never know the details, but at a stroke Caldera has acquired a huge war chest.

And now they've filed for an initial public offering. Interesting times ...

Old Nerd's Almanac

Prognosticating about the future of Linux is a tough call. Certainly five years ago nobody expected it to be where it is now. Is it possible to see where things might be by the end of 2000, though?

Peering into my crystal ball, certain trends are fairly clearly emerging. Linux is fragmenting into three or more distinct segments, and these are going to be the scene of vicious fighting between companies who see them as useful markets.

The first and best established sector is the small to medium sized departmental server; the Red Hat box in the corner chugging away serving web pages, or the gateway machine providing ISDN or ADSL dialup and firewall services to a small company office. This market is dominated by Red Hat and will probably remain so. (Caldera are focussing on the desktop, to the detriment of their distribution when it is viewed as a server; some useful daemons such as atalkd are missing and the whole slant of their system is towards a slick mouse-user's experience rather than providing tools like php3 and mod_perl compiled into Apache.) Red Hat seem to be focussing on the supported departmental server market with their Professional and Deluxe packages. They will have competition from SuSE (especially in Europe), TurboLinux (in Japan and the Pacific Rim), and less-well-known but potentially superior offerings such as Rock Linux or Mastodon (both of which expect to be installed and run by a UNIX guru and in the right hands deliver superior performance).

The second market segment is the desktop. I'm going to ruthlessly ignore the hobbyist/student user base -- people who, to be fair, I feel a lot of affinity for -- because they aren't where the money is. The real desktop market is the cheap, 500-pound PC bought by someone who probably never ever installed a software package in their life. Corell have made a spectacular bid for dominance by shipping copies of their easy-to-install distribution with a whole range of cheap motherboards, thus ensuring that low-cost PC's assembled in small shops will come with at least one operating system; expect Corel to capitalize on this with support sales of Word Perfect Office 2000 (entering beta test in December 1999). But they won't have it all their own way. Caldera has a better-established desktop environment and is able to provide StarOffice instead of Word Perfect Office; they also have better DOS and NetWare integration (by way of their relationship with Novell) and may be able to capitalize on their early lead. Red Hat are still in this field but it's not their strong point -- the graphical installer introduced in Red Hat 6.1 is helpful, but not enough on its own. The level of expertise necessary to set up and run a Linux desktop system is likely to fall relentlessly as ease-of-use and integration become the goals of the companies building the distributions.

The real loser in all this is, of course, Applix. There's a multi-way contest for the desktop office suite; StarOffice (backed by Sun) in one corner, Word Perfect Office in another, the freeware solutions (AbiWord, KOffice) trying to catch up ... and ApplixWare, costing more than any of the others and with less of a market lead. I'd therefore be very unsurprised to see Applix bought by one of the bigger distributors within the next year and their core product given away for free in an attempt to expand in the critical desktop segment (much as StarDivision was bought by Sun in an attempt to provide a solution to the MS Office requirement for their users).

All this commercial stuff is boring, though. It's not where the real action is.

To start with, the GNOME/KDE war will heat up. KDE 1.89 is out just about as you read this article; KDE 2.0 should be out by summer, and catches up with GNOME's technological lead by adding serious CORBA support, an office suite, cool development tools, and a new (and powerful!) web browser/desktop. But GNOME isn't standing still. By Summer 2000 Mozilla will be out. Mozilla is still pre-alpha, but already looks incredibly cool; this is one web browser that is quite likely to give rise to "back from the grave" headlines when the mainstream computer press notices it. XML-based from the outset, with supports for all the style sheet standards, plus javascript and java and the usual extras, Mozilla is a complete re-write of the original Netscape browser. And by the time it ships it should be fully GNOMEified.

And as if that isn't enough, Magellan should surface early in the year. Magellan is likely to cause lots of sleepless nights at Lotus, because it looks like being a Lotus Notes killer for Linux. It's a combination PIM, mail user agent, and general messaging system; although the first incarnation will be basically a client tool, the server component should separate out fairly rapidly. This will give Linux clusters almost all the capabilities of Lotus Notes without the need to shell out big bucks for a Domino license (and yes, Lotus Domino -- the Notes-via-HTTP server -- is already available on Linux). Linux is already an incredibly powerful middleware/groupware platform. With tools like Magellan and Mozilla appearing, it can only get better.

The Linux kernel isn't getting feebler, either. 2.4 should ship before the end of the first quarter, providing integrated PCMCIA support, USB, IRdA, and various stability improvements over 2.2. More to the point, 2.4 should eventually incorporate the ext3 journaling filesystem, SGI's XFS filesystem, raw device access (essential if high-end database servers such as Oracle 8i -- already available on Linux -- are to deliver maximum performance), and a capability-based security mechanism. This latter item is very important. Linux is already partially capability based, but when 2.4 is out and the tools to control it are available it will move all the way towards this next-generation security model. Security is not something a server operating system can ignore; the traditional UNIX user/group ID system simply isn't up to the job, but with capability support it will be possible to delegate "fine grained" permissions to users or daemons that need it. No more need for a root user ID that can destroy the universe if it sneezes at the wrong time.

A second item that should show up in 2.4 is clustering support. Clustering is already with us to some extent, but by the time 2.4 matures there should be some general agreement about how to do clustering. Big clusters of multi-processor Linux servers will be catching on; one US government agency has already ordered a supercomputer from Cray that will be delivered in 2002. This supercomputer should have in excess of 1024 IA64 (Itanium) processors ... and will be running a version of Linux developed by SGI. With growing penetration into the supercomputer and mainframe markets, expect to see a strong emphasis on clustering and symmetric multi-processor support in Linux kernel development through 2001.

The third big thing to expect in the kernel is a re-design to cope better with hot-swappable peripherals. PCMCIA cards are all very well, but it also helps to be able to yank and re-insert things like USB devices -- to say nothing of FireWire and PCI bus devices. To do this properly will require a redesign of some fundamental kernel structures; expect to see this surfacing in a development stream that will probably cough up something called Linux 3.0 some time in 2001-2002.

And then there are those market segments again. Supercomputers may be sexy but they don't move mountains on their own. But at the bottom end, there are signs of Linux running on toasters. A "toaster" in this context is an embedded appliance; a set-top box, or a thin server -- a server that is a black box designed to pump out web pages -- or an in-car MP3 player. The embedded system market doesn't have a lot to fear from Linux yet, but Linux is probably more attractive for embedded applications than, say, Windows NT. Moreover, it can be skinned down; the small cut-down ELKS kernel will happily run on a Psion 3a with only 512K of RAM. Most of the embedded systems based on Linux through 2000 will probably be thin servers; things like the Cobalt Qube or Rebel.com NetWinder. But I expect some radical surprises, like the cellular data-enabled touch-screen device Nokia demonstrated in 1999. It is not beyond the realm of possibility for Linux to mount a serious challenge to Symbian in the cellphone operating system market, at some time in the future.

I'm not saying much about commercial applications development, you'll note. That's because it will happen, whatever I say. The availability of powerful, free development tools such as Glade, KDE Studio and DDD is driving a transformation in the ease of producing graphical applications on UNIX. With commercial offerings such as CodeWarrior and Delphi showing up in 2000, the oft-heard complaint that Linux development uses primitive (read: command-line) tools will ring hollow. The only conspicuous absence will be Microsoft; even if they suddenly decided to make a U-turn on Linux, it will take at least six months for something like Microsoft Office 2000 for Linux to show up.

Finally, there's scope for some major surprises. I don't expect, but wouldn't be too astonished, to see major upheavals hit the field. The Berlin Consortium might deliver a graphics environment that knocks X11 into a cocked hat -- as they've been promising for 18 months -- and if it has some kind of "compatability box" for running old X applications in it could catch on rapidly. Some time in 2000, TransMeta Corporation should be delivering CPU's in bulk, and a Linux link isn't out of the question (given that TransMeta's chief scientist is one Linus Torvalds). Nasty surprises as well as pleasant ones are possible. For example, US patent law is such a mess that a "submarine" patent that renders Linux unusable could surface. It would be fought in court, of course, and today companies like Red Hat or Cobalt Networks can afford sizeable legal fighting funds, but it will be unpleasant if it happens.

One possible flash-point is DVD support. The DVD consortium insist on charging money and obtaining non-disclosure agreements from people who want access to their standards. A group of hackers called DeCSS, annoyed at their inability to view DVD movies on Linux, reverse-engineered the CSS encryption scheme used by DVD; only to be accused of piracy, rather than a legitimate desire to watch the media. Linux may be perceived as a threat by the entertainment media because its openness makes it possible to crack virtually any copy-protection scheme if a player is available on that platform. (For example, a Real Audio crack surfaced some months ago; it relied on grabbing the raw bitstream going into the sound board, after the player was finished decoding it. Arguably, this sort of hack is a lot harder to carry out on an operating system where you don't have access to the kernel sources and the ability to re-write bits of it.)

The worst surprise that could happen to Linux in 2000 is a new distribution. The prospect of Microsoft Linux is one that must keep directors of companies like Red Hat and Caldera awake at night; Microsoft have the muscle to take over the entire Linux field, should they decide to do so, and once they own it they could do just about anything. (For example, roll out a new kernel -- written in-house -- and call it their own, then use "embrace and extend" to make it impossible to run a Linux system in business without it.) Luckily, Microsoft are tied up in a lengthy appeals process and will have to watch what they do very carefully. It was exactly this sort of legal oversight that hamstrung IBM during the 1970's and 1980's; an attempt by Microsoft to cripple their perceived opposition at this juncture would almost certainly have very costly side-effects. So the likelihood of a serious attack coming from Microsoft in 2000 is low.

[ Site Index] [ Linux Index] [ Feedback ]