| Tuesday, January 16, 2007 |
14:26 - Engineering for the Future
http://hivelogic.com/articles/2007/01/14/regarding-the-iphone
|
(top) |
Dan Benjamin of Hivelogic has an interesting take on the iPhone and its technical prospects. A lot of its thesis focuses on OS X and its UNIX pedigree, and while it's basically correct, I would be remiss if I didn't quote the following and expand on it pointlessly:
To really understand why OS X on Intel (and now on the iPhone’s CPU) isn’t such a big deal, we need to consider that at the core, OS X is UNIX. It’s based on FreeBSD, a great UNIX system built (primarily) for Intel CPUs and known for rock-solid stability. This is a big part of why Macs crash so rarely these days.
You’ll notice that I mentioned (above) that FreeBSD was developed for Intel. Later, it was “ported” to work with other CPU’s as well, but its primary focus has always been Intel. I believe that this focus played a big role in why Apple selected it as the basis for OS X. I think that Apple had been planning a switch to Intel all along, biding their time. The fact that the Apple had a complete version of OS X on Intel since day one is proof of this. So is the existence of Darwin (Apple’s open source version of the OS X core operating system) for Intel.
In fact, Apple had to port FreeBSD to PowerPC before they could really get started working on OS X. It’s even possible that Mac OS X 10.0 was an Intel operating system before it ran on the existing Mac hardware of the time.
Not only is it possible, it's historical fact.
Mac OS X is not, strictly speaking, "based on FreeBSD". Nor is Darwin, really. What they are is based on OPENSTEP, the successor to NeXTSTEP, both of which were essentially wacky polygamous marriages of the BSD userland, the Mach kernel, and the NeXTSTEP/OPENSTEP libraries (now called Cocoa, and the reason why all the functions in Cocoa have "NS" prefixed to their names).
OPENSTEP was always Intel-native, though versions for SPARC, 68K, and HP PA-RISC also existed. When Apple bought NeXT, it inherited OPENSTEP in its entirety, and Mac OS X isn't so much a whole-cloth redesign of classic Mac OS as it is simply OPENSTEP with a new display layer (Quartz), a new presentation interface (Aqua), and a tacked-on Mac OS compatibility layer (Carbon). Everything else is a direct descendant of the NeXT stuff. But he's right in that the PPC and Intel versions of OS X have existed in parallel for as long as Mac OS X has existed—more than that, they've both been there since 1989.
The relationship FreeBSD has with Mac OS X, as I understand it, is that the userland structure, libraries, and code for the system binaries (which are not cross-executable) come from FreeBSD, and keep getting infused into Mac OS X in periodic merges to keep it current. (Before FreeBSD was created in the mid-90s, OPENSTEP and NeXTSTEP inherited their userlands from 4.3BSD and 4.4BSD Lite; after those branches stagnated, FreeBSD became the de facto authoritative BSD source.) But the Mach kernel and the Cocoa libraries have always been non-BSD-specific, as have a lot of the additions to the basic BSD architecture that NeXT and Apple have engineered. If you look at the UNIX Timeline at http://www.levenez.com/unix/, you'll see how closely FreeBSD and NeXTSTEP/OPENSTEP/Mac OS X are intertwined—pretty closely, but not so closely as represented in Benjamin's article.
So the upshot is that it isn't so much that Apple selected FreeBSD to be the basis of Mac OS X. It's more like NeXT selected a BSD to be a basis of NeXTSTEP, and Apple inherited that decision ten years later.
Now, of course, none of that is to diminish the significance of the overall point, which is that OS X has always had an architecture-independent streak a mile wide. Indeed, that was the whole point of OPENSTEP—it ran on everything from SPARC-based Sun boxes to Windows NT. Similarly, Cocoa and Xcode today are platform-agnostic abstractions. But the fact that you can compile a Cocoa app for Intel or PowerPC by clicking a single checkbox at compile time is not something we ought to regard as a recent innovation; rather, it's something that's been inherent in the NeXT world since day one. That's the real core of Steve Jobs' ideal world, as illustrated by how he designed NeXT. It may well be that his designs on the world of technology still mirror what he planned with NeXT: that while whole-widget engineering is still the crown jewel, it doesn't preclude the software platform from being agnostic of the hardware it runs on, so that it can take advantage of whatever gear is most suited to the task at hand.
The real key to understanding OS X—and how ostensibly quickly it came to be—is realizing just how much of the work was done back in the early 90s, in a little-known backwater software and workstation concern, laying the groundwork for some then-unguessed future company that would one day reap all the benefits of NeXT's development efforts and deploy them in everything from servers to laptops to, well, set-top boxes and phones.
How proud Steve must be.
|
|
Brian Tiemann
