LWN's Obviously Incorrect 2004 Predictions

The new year is upon us, and so, like many other publications, we feel an irrational urge to wave our hands in the air and make predictions for what we think the coming year will bring us. So here we go. Needless to say, anybody who is thinking of acting on any of the following would be well advised to get a second opinion...

Enterprise Linux

The "enterprise Linux" business came into its own in 2003, as Red Hat, in particular, found a steady stream of willing customers which drove the company into a profitable state. Red Hat's enterprise offerings must be providing value to the company's customers, given the claimed 90% renewal rate on enterprise support contracts. But the per-system licensing of Red Hat Enterprise Linux has rubbed some community members the wrong way; many developers feel that Red Hat's contracts do not reflect the sort of world they thought they were helping to build.

So, we predict that, in 2004, the enterprise Linux backlash will grow, and we will begin to see whether that backlash can change the enterprise Linux market. A number of free enterprise Linux projects are out there, including CaOS, Whitebox Linux, and UserLinux. These projects have an uphill road ahead of them; to be successful, they will have to convince skeptical companies that they will be able to provide high-quality support for many years into the future. They will also have to make independent vendors see them as important enough to certify applications for. Oh, and, of course, they will have to create a top-quality distribution aimed at the needs of this sort of customer. Creating that distribution will not be easy, but it may prove to be the simplest of the challenges faced by any would-be challenger to Red Hat's enterprise offerings.

The interesting thing is, of course, that these challenges look remarkably similar to those faced by Linux itself a few years ago, when the idea that Linux could pull the rug out from under proprietary Unix systems and challenge Microsoft seemed ludicrous to many. But it happened. Now, challenging enterprise Linux with truly free Linux looks like a daunting task. But it may yet be that a free distribution combined with a distributed network of supporters could supplant today's enterprise offerings in just the same way that Linux has taken Unix's place. The community is capable of amazing things. Not everybody would see such an event as a good thing, but it could happen.

Desktop Linux

The desktop wars will heat up again in 2004. For those of us who remember the KDE/GNOME flame wars of years past, the relative calm and cooperation which has prevailed more recently has been a welcome thing. But there are pressures building which threaten to upset the peace.

The first of those pressures is licensing. Ironically, KDE's choice of the GPL for its libraries may work against it here. The looser GNOME library licensing allows its toolkits to be used, royalty-free, with proprietary applications. Proprietary KDE applications can only be distributed by paying royalties to Trolltech, which owns the Qt libraries. Many users and developers would rather not see proprietary applications exist at all, or, at least, not without paying those who have developed the underlying toolkit. These people are happy with KDE's licensing. Most users, of course, don't care. Distributors, however, usually want to enable vendors to sell applications on their platforms. This interest will push them toward the GNOME camp.

The other point, however, is that distributors are increasingly under pressure to make a choice. Supporting two desktop systems adds to the total workload of maintaining a distribution, and that costs money which may not be available. There is a common perception at this point that the two desktops are functionally identical in all the ways that matter; if that is true, why bother with two of them?

In 2004, these pressures will lead to rising emotions in the camps of both desktops as they see decisions being made for or against them. Perhaps the result will be a greater degree of cooperation between the two development communities via freedesktop.org or other mechanisms. Or, perhaps, our newsgroups, web sites, and mailing lists will once again play host to heated debates and flame wars in vain attempts to establish one desktop as being superior to the other.

Beyond that, however, the hackers will stay busy and desktop Linux will amaze us again. In 2003, it was widely recognized that Desktop Linux has everything that many, if not most, business users need to get their jobs done. 2004 will be the year that desktop Linux stops playing catch-up (in some areas, at least) and truly begins to blaze interesting trails of its own. Projects like Dashboard, GNOME Storage, and Reiser4 are just the beginning of a wave of innovative projects which will change how we use our computers.

2004 may not, however, bring Linux into many more homes. A Linux system is more than adequate for Grandma to send email and wander around on the web. Your editor insists that his children use Linux for their email, browsing, and homework tasks, and it handles those jobs well. The sad truth, however, is that there still needs to be a Windows system around for other vital tasks - such as playing games. Home users are not interested in dual-boot systems; until Linux can do everything they need, they will stick with the same old stuff. Linux may eventually have a free application base which replaces many of the commercial offerings currently filling the shelves of computer stores, but it remains hard to imagine free games, for example, which can compete with the hit-driven commercial variety. Until there is a lively market for commercial Linux applications, there will be some hard limits to how many desktops we can occupy.

Legal issues

The SCO case will drag on, and become more complicated, in 2004. IBM may well succeed in getting many of SCO's complaints dismissed early in the year, but SCO probably has a good chance of keeping some of its breach of contract charges alive. SCO may have to retreat to some of its earliest charges (i.e. JFS, RCU, NUMA, SMP), but IBM may have to go to trial to prove that its code in those areas is not derived from SCO's Unix. SCO can probably muddy the waters enough to keep the judge from dismissing the case outright.

Even if the IBM case is dismissed in 2004, however, there is the issue of SCO's threats of copyright infringement suits against Linux users. One may be tempted to dismiss these threats as just that much more empty SCO bluster. It is worth considering, however, the pressures that SCO will be under, including the agenda of its lawyers and the looming "dividend" payments on the BayStar investment. SCO has no hopes for increasing revenue from its remaining software products at this point; it must litigate further to bring in cash. With the lawyers in charge, chances are that SCO will, indeed, launch new suits.

In fact, the company may well find backbone-challenged Linux users that will cave in and pay up rather than risk a court battle. Such an event will do short-term wonders for SCO's stock price and cash flow.

The simple fact is, however, that the SCO Group has still put forward very little evidence to back up its claim, and what evidence it has presented has mostly been laughed off the stage. The company's claims to own the "Unix ABI" will get no further. Beyond that, Novell's new copyright assertions have the potential to stop the show dead, at least until that dispute has its own day in court. But, regardless of the validity of Novell's claims, SCO's case is empty and the world, increasingly, is seeing that. By the end of 2004, the SCO cases will probably still be alive in some form, but the end will be in sight.

As an aside, Novell will face a severe test of its credibility in the eyes of the community. Nobody wants to see the SCO case resurrected in the future by a Novell which, perhaps after a management change or two, decides that its Unix copyrights (if they are Novell's) might yet be worth something. If Novell is serious about being a part of the Linux community, it needs to make a statement, soon, about just what it intends to do with the Unix copyrights it claims to own.

The GPL may have its day in court. The SCO Group has, of course, stated its intent to break the GPL in court. But that company's arguments, thus far, have failed to impress. SCO's GPL challenges should not get far. More interesting GPL-oriented cases may come from a different direction.

Many developers working in the industry are full of stories of rampant GPL violations, especially where embedded systems are involved. Last year's episode with the LinkSys WRT54G router is just the small tip of a large iceberg in this regard. To an extent, people have been willing to look the other way; it just hasn't seemed worth the trouble to challenge closed-source uses of GPL-licensed code in many cases. There are developers, however, who are increasingly unwilling to close their eyes to violations of their licenses. Expect more challenges against vendors using GPL-licensed code in non-licensed ways. The lack of any court decisions on the GPL will eventually embolden a violator to try his luck in front of a judge. At that point, we will begin to see what the judicial system really thinks of the GPL.

Security

2004 will make us care more about security. In 2003, we saw an ominous string of attacks against the servers which support the Linux development community. There is no reason to believe that these attacks will stop anytime soon. Sooner or later, perhaps in 2004, somebody is going to do some real damage on a scale we have not yet seen. A major breach, perhaps compromising the systems of many Linux users, will cost us money, time, and much credibility.

In recent years, most attacks against Linux systems have exploited known vulnerabilities for which patches existed. A well-managed site is nearly immune to attacks using known vulnerabilities; all of the major distributors are quite good (usually) about quickly preparing updates when a problem comes to light. The attacks we saw at the end of 2003, however, made use of previously unknown holes in rsync and the kernel. Defending against unknown vulnerabilities is much harder, and there do exist attackers who realize this, and who are smart enough to find such problems. In the coming year, we may well see some truly scary exploits of this sort of "zero-day" vulnerability.

There is some good news, however. By the end of 2004, we will see wider deployment of hardened Linux systems. The incorporation of SELinux and various other security technologies into the Fedora Core distribution (and, from there, into Red Hat Enterprise Linux) will drive much of this deployment, and threats from the outside will do the rest. Adding SELinux is a significant step in the evolution of Linux distributions; if this work is done properly, Linux users should soon have a much higher level of security available with a default system install. Proper containment of security breaches should, for example, make that next web server buffer overflow be much less of a threat than it is now.

Kernel

2.7 kernel development will begin after the 2.6.0 kernel has had a few months to stabilize. Expect the 2.7 development series to be quite different from 2.5, however. By the time that 2.5.0 came out, there was a massive backlog of patches waiting for inclusion. The 2.4 stabilization process had taken nearly a year, and there was a long shopping list of planned changes for 2.5, including the block layer rewrite, expanding the dev_t device number type, a new loadable module subsystem, a new kernel build mechanism, asynchronous and direct I/O, and many others.

On the eve of 2.7, the "patch pressure" is far lower. There's no end of ideas for 2.7, including virtual memory work (page clustering, shareable page tables, etc), the never-ending desktop interactivity work, and much internal reworking to eliminate race conditions, and so on. But many users are (or will be) far happier with 2.6 than they were with 2.4, and the list of features that the Linux kernel must have to not be jealous of its Unix predecessors is shrinking. The Linux kernel is maturing, in other words. It may well be that, with 2.7, the pace of change begins to slow a little.

Or maybe the kernel hackers will come up with some amazing new ideas and run with them; at that point, all bets are off.

To conclude...

It's going to be an interesting year. That, perhaps, is the only truly safe prediction to be found among all the others on this page. All the rest are offered with no warranty as to their veracity, suitability for any particular purpose, or connection with any sort of reality whatsoever. LWN.net does not provide indemnification for users of its predictions - though purchasers of our "predictions license" (available for a limited-time special half-price deal through January, 2007) will get a promise from us to not sue them.

---

to post comments