12/9/2003 -- The holiday season is a time for reflection on the year just past; but also for wanton speculation about the year(s) to come. With only that meager justification, here is a list of what I see as the most interesting (but not always the most important) trends and/or events in the Linux/Unix world this year.

Increasing Commodification of Hardware (and Software?)
As you might gather from the amount of text I've devoted to it, this is a good contender for the most significant development in Linux news -- indeed, in general computing -- this year. The slumping American economy had at least one benefit, forcing vendors to drop hardware costs sharply this year. Not least because of memories of the outlandish IT budgets during the Internet bubble, standardization on commodity hardware will be the watchword for 2004, according to recent IDC predictions -- specifically, servers based on x86 chips. Whether Linux or Windows will predominate on these servers remains unclear. As Sun's financial woes highlight, companies focused on other, more specialized architectures will face continued losses in that market. The costs of specialized component production -- for instance, RISC microprocessors such as the Sparc -- are simply too high to sustain in face of Intel's consumer CPU juggernaut. While it will take years for the effects of this shift to become clear, Linux is expected to comprise 10 percent of all server sales in 2004 -- what IDC calls a "tipping point."

The tight economy has also brought software costs into the harsh light of day, where Linux looks more attractive. Even Oracle now runs on Linux, according to Larry Ellison. Part of Linux's difficulty in making inroads into business was the perception, difficult to deny until recently, that it was "unsupported" or required arcane knowledge to get it installed and operational. Dell's amazingly successful model of server construction, where customers sacrifice component choice in favor of ease of installation, has eliminated this problem by putting the onus of operating system installation back on the manufacturer. IBM, HP and Dell -- the three major players in the Linux server space -- now all offer successful rack-mountable servers with Linux preinstalled, which they're able to do by using a known, thoroughly tested set of components. While consumer choice is restricted by this tactic, in another sense, it's another benefit of the increasing commoditization of hardware. As hardware becomes cheaper and cheaper, it no longer matters what type of hardware the server is comprised of. Would you spend lots of time worrying about who made the heating element in your toaster? Probably not-and so, says industry, should we treat servers and operating systems. Only the support (or, you could argue a la Naomi Klein, the brand) matters.

Increased Corporatization of Linux
Dell, IBM and HP remain the key players in the Linux server market, which IDC reports increased 40 percent in the second quarter of 2002. Two weeks ago, Sun jumped into the fray with the announcement of their Java Desktop System -- a repackaged SuSE Linux installation with StarOffice and other applications integrated. Sadly, increased vendor support for Linux doesn't mean increased reciprocation with the Open Source community. RedHat's recent decision to end support for RedHat Linux in favor of its enterprise product (RedHat Enterprise Linux) is unforgivable to some, beneath notice to others. While numerous other Linux distributions are available, RedHat's capitulation merits close observation. Is this a harbinger of further "defections" from the major Linux distributors? The Linux distribution landscape is a swamp from the CIO's vantage, in which only two or three distributions (SuSE, RedHat) have any real corporate clout. It's undeniable that Linux has and can continue to benefit from business's involvement. Donations by business of no-longer-profitable technologies to the Linux community, such as SGI did with the XFS filesystem, are one example-and handouts or not, such donations have real value. But the question is: how will increased adoption of Linux by industry heavyweights affect the Linux community? Will Linux fragment even more, into "enterprise" and "end-user" camps? How will the first serious legal challenge to the GPL play out? Does SCO's lawsuit even qualify as "serious?" When you answer, remember that the Warren court is no longer in session.

Increasingly Vicious Copyright and Patent Battles
Kazaa notwithstanding, copyright is alive and well. Hard times have caused many companies to step up their defense of intellectual property rights. SCO's much-ridiculed legal battle with IBM over its ownership of Unix, widely seen as an attempt to drive up its stock price, is but one example. Microsoft recently announced that it would be charging a license fee for use of the FAT filesystem, which is covered by several patents. In a legal climate where someone was able to copyright the "Happy Birthday" song despite its being composed in the 19th century, we can expect these sorts of tactics to continue, and probably succeed, even in the face of widespread, perhaps tolerated civil disobedience.

Macintosh OS X
Macintosh, ever since branding itself the stylish alternative to the thought police in 1984's "Big Brother" commercial, has lacked something in every department except style -- until now. Macintosh OS X, a full-fledged BSD-derived Unix system, continues to win new converts among Unix and Linux zealots, despite backlash among long-time Macintosh users at replacement of the tightly-integrated form and function of the original Macintosh Finder (the Mac's desktop engine) by a cobbled-together version of the NeXTStep interface. The inferiority of the OS X interface to the old Macintosh System 7/8/9 interface is a testament to the ingenuity of the original Mac's designers; but its superiority to anything else on the market is a not-so-subtle rebuke to the much-imitated and -criticized Windows interface, which many Linux desktop distributions take as their model: Here's how it can and should be done. There's even an entire O'Reilly site devoted to the new operating system: surely a token of respectability.

Increased Doomsaying
The chorus has been sounding the death knell for Moore's law for some time. This year, however, Intel's own researchers released a paper ("Limits to Binary Logic Switch Scaling," in the November issue of the Proceedings of the IEEE), predicting the end of the halcyon years. At current rates of development, chips will hit a 5-nanometer limit around two decades from now. Transistors with gates of that miniscule width become unreliable, because quantum effects can cause electrons to jump the gaps. Speculation abounds as to exactly when this will occur, and about alternative technologies that offer some promise of extending the longevity of Moore's law (diamond wafers, quantum computers, and so on), but consensus seems to be emerging that glory days of traditional silicon computing, like the steam engine, might be almost past; we may now be on the downside of the curve. Regardless of the specifics, any slowing in Moore's law will have the side benefit of encouraging the development of software that can use existing processors more efficiently; nor are the prospects for alternatives to silicon as a substrate exhausted. Either way, you can bet that if Intel (whose founder Gordon Moore coined the "law" that bears his name) is thinking seriously about it, it's an issue.

Near-Limitless Storage
The truest adherent to Moore's law isn't the processor -- if you look closely, you'll notice that even Intel has amended Moore's original prediction to a vague doubling "every couple of years" -- but storage. In a recent interview, James Gray, 1998 winner of the ACM Turing Award, notes that disk capacities have been increasing at approximately 100 percent per year since 1989, while disk access times have been increasing at a paltry 10 percent per year. In short, it's soon going to be possible to buy more storage than you can hope to reasonably retrieve. At this point, everything changes -- backup strategies, programming strategies, you name it. The disk becomes a sequential device, argues Gray, which can store every version of a file since its creation, and need never overwrite itself. Moreover, if iSCSI and other storage technologies are incorporated into the disk drive itself, as Gray argues is inevitable, then storage becomes abstracted from the computer itself, and everything depends on IP. These developments may seem distant now, but terabyte drives are predicted within a few years (two years, according to Seagate); certainly by the end of the decade.

What's Next?
To many industry pundits, Bill Joy's departure from Sun is the darkness at the end of the tunnel for the company. There's perhaps more than a little truth in this. A revealing interview with Joy in Wired wins my vote for most dismissive offhand comment of the year. In reference to Linux: "Re-implementing what I designed in 1979 is not interesting to me personally." Though he relegates the entire Linux community to the status of kids playing with toys with that comment, in context, the quip takes on even more weight. What he's saying is that traditional computing is, in many senses, out of ideas. Faster drives, faster chips and remixed operating systems from the 70s probably do not comprise the computing environment for the next 25 years. The Unix platforms of today owe their heritage to technology developed several decades ago; what the next 25 years will bring is anyone's guess, but it does look as though some fundamental limits are being reached. In many ways, this is a good thing for IT as a whole -- the new ground broken by research and development, and the new applications derived from new methods of computing, can't help but ensure that IT will be a rewarding (and lucrative?) career for years to come.

What's your take on these issues? Post your comments below!