Apple: Unix for People, Unix for the Masses
by Timothy Prickett Morgan
Here's a riddle: What has all of the user friendliness of Windows, the openness of Linux, the stability of Unix, and the inexpensiveness of the X86/X64 platform all rolled into one? The answer is this: The future Apple desktop, laptop, and server product lines, provided Apple doesn't somehow manage to screw up its third major product transformation. As the entire planet knows, Apple is moving to Intel chips and off PowerPC chips, and now is a good time for the company to re-examine its business model.
Apple probably will not completely re-examine what it is doing, of course, because Steve Jobs has always done exactly and only exactly what he wants to do--even if it does get him fired sometimes. But hope springs eternal, particularly among journalists who like computers, so there is hope.
First and foremost, kudos to Apple for picking BSD Unix as the core of Max OS X. In six week's time, Apple has shipped over 2 million copies of its "Tiger" OS X (technically, release 10.4) operating system, which was launched with much fanfare on April 29, becoming (once again) the volume Unix operating supplier in the world. It is hard to say how many copies of the various open source BSDs are distributed, since the projects controlling them and their site mirrors do not keep track, so the volume award goes to Apple.
I have made a lot of noise about how Sun Microsystems has shipped 1.3 million freebies of its Solaris 10 Unix platform from the end of January through the first week of May, and that this is an absolutely huge installed base compared to commercial Linux shipments from Red Hat and Novell, which together ship about a quarter million units in 13 weeks. Solaris has nothing on Tiger OS X. People have paid for 2 million copies of Tiger, at a rate that is three times higher than the starting rate for Solaris 10 in February of this year. (Imagine if it were free and open source!) To put it bluntly, Tiger is the volume Unix platform, and that is without the shift to Intel chips from PowerPC chips. At $129 for the desktop version and between $499 and $999 for the server version, Tiger OS X has probably brought Apple about $275 million in six weeks. To make about the same amount of money under its free usage/fee services pricing scheme, Sun will have to pump out about 2.5 million downloads (which it should do by January 2006) and get at least 20 percent of these customers to pay out for for-fee technical support. Obviously, Sun is being optimistic and courageous in its approach. We'll see if it works out. (I am optimistic that it will, especially considering the very reasonable prices Sun is charging for Solaris 10 support.) In his keynote address, Jobs said Apple had hoped to have about half of the Mac installed base on Tiger within a year; Tiger comprises about 16 percent of installed machines right now.
While Apple has made the jump to Intel chips for future Macs and Xserves--the initial Apple boxed based on Intel processors are due by June 2006, and Apple is promising to have the whole product line transitioned to X86/X64 processors by the end of 2007--the company has not said that it will supply a version of Mac OS X on other Intel-based machines. But clearly this is an option since there is no reason why the BSD-based Mac OS X cannot be made to run natively on any modern X64 box--and that almost certainly includes Opteron-based machines, by the way. (More on this in a minute.)
The nonsense that Apple is talking about the move to Intel processors being predicated on the long-term roadmap of the Pentium and Xeon processors offering more power than the roadmap of IBM's Power line of processors is just that: nonsense. What can honestly be said is that the three original partners in the PowerPC Alliance from 1991--Apple, IBM, and Motorola--each brought their own needs and incompatibilities to the table, and they never hammered out their differences. Motorola was mysteriously put in charge of creating the 64-bit PowerPC chips. (Remember the PowerPC 620 and 630? Of course not, they never saw the light of day) IBM was put in charge of making some 32-bit designs and both IBM and Motorola ended up competing for the embedded controller market (once ruled by Motorola's 68K series of chips, which were also at the heart of Apple Macs) and for Apple's business, too. Rather than working together to make a single set of 32-bit and 64-bit processors that allowed Mac OS, OS/400, AIX, Windows, and any other operating system to run on these desktop, workstation, and server machines, the PowerPC Alliance splintered technically long before it fell apart economically or politically. And that, ultimately, was its downfall. So why is Apple moving to Intel chips? Well, if Apple cannot control its destiny in terms of chips, it will at least be in the same boat as the rest of the industry, and not playing fifth fiddle to the IBM's iSeries and pSeries servers, Sony PlayStation 3s, Xbox 360s, and whatever other dozens of other vendors of electronics who use PowerPC chips make and sell. With only 3 million units a year, Apple could not make IBM alter its roadmap, and Freescale Semiconductor, having been summarily orphaned by Motorola, is not about to take any huge risks by substantially changing its PowerPC roadmap.
The sad truth is that IBM and Apple should have long since ported Mac OS to the Power-based server line created by Big Blue, and IBM should have listened to Apple and created a low-powered, 64-bit PowerPC chip that could run Mac OS X in a laptop without cooking a user's legs. IBM most certainly could have done this, but it has had other priorities--like ramping up performance on the Power5 chips as much as possible to compete in the Unix and proprietary midrange and enterprise server space or selling low-powered chips for embedded devices. IBM's PowerPC 970 and its supposed kicker, the PowerPC 970MP with dual cores, was a high volume product in relation to Power5, but it probably didn't make IBM as much money or Big Blue would have fought to retain the Apple business. For all we know, IBM made such promises. It doesn't matter. This should have happened in 1995.
No matter. If Apple wasn't planning on moving to the Intel architecture when it created Mac OS X, it certainly laid the groundwork for such a move by using BSD Unix as the heart of the operating system. The FreeBSD Project contributors I talked recently to say that Mac OS X will run fine on just about any X86 machine (there is, of course, the driver issue for peripherals), and moreover, that it was designed to do so. Apple, having been burned so many times by Motorola and IBM, and not wanting to be burned in the future by Intel, has kept its options open. You can bet that Apple and AMD have been talking to each other, too. (If I were Jobs, I would have talked to AMD before I talked to Intel, in fact, just to get the best possible deal from Intel.) By having the operating system based on BSD, whole chunks of it are largely portable, just like the open source BSDs are.
So why will it take 18 months to roll out the Intel-based Apple machines? Because Apple thinks it is a hardware manufacturer, and it is in love with the idea of designing and building computers. And that is fair enough. Let's face it: Apple has the sexiest computers on the market, whether they are desktops or iPods or xServes. But if Apple is really interested in taking the X86/X64 market by storm, it may be time to let Mac OS X go--and really let it go. At the very least, Apple might be smart to create an open source community and let that community do a port of all the relevant pieces of Mac OS X to all kinds of X86 and X64 machines. For native Intel code, this would be a great strategy.
The stickler, of course, is the Rosetta emulation environment, which is largely based on the QuickTransit emulation software developed by Transitive and which is certainly not open source and is wickedly expensive. So an open source Mac OS would not be able to support emulated PowerPC binary applications. So, OK. This is perfect. If you want a sexy machine that can run emulated PowerPC applications as well as native X86 code, you have to buy the machine from Apple, and if you want to start from scratch, Apple can sell you a slightly cheaper X86 box or you can go to the open source community, get the open source Mac OS, and pay Apple for support. Such a model might drive a lot more volumes than waiting around for Apple to get its X86 boxes designed and built. And it might build a much larger and more vibrant Mac OS ecosystem.
If Sun is right about one thing, it is that the Solaris ecosystem is what matters--not Sparc hardware. Apple would probably do better if it learns that lesson now and focuses on making a great Unix platform that is suitable not just for nerds who love Unix, but for people who don't want to learn Linux all at once (and maybe can learn Unix in smaller bites) and who are frustrated with Windows. That is a vast market--much larger than playing to the Mac installed base. If I had the option of choosing between Mac OS, Linux, and Windows, the 15-year-old in me who fell in love with an Apple IIe that ran the first flight simulator I ever saw knows which is the better--and cooler--option.