Universal Applications

How the transition to Intel is very different than the move to PowerPC

As noted in Why Apple Hasn't used Intel Processors Before, the Classic Mac OS was closely tied to the Motorola 68k processor. In the mid 90s, Apple's move to PowerPC hinged on the ability of PowerPC chips to emulate 68k code. That emulation burden was a necessary evil because so much Mac software, including Apple's own, was tied to a processor-dependent foundation.

There was no other way to get past this. PowerPC offered the only clear upgrade path; the newest 68k processor, the 68060, was a disappointment, and the inability of Intel's current chips to emulate 68k code fast enough prevented Apple from building Pentium Macs.

That processor dependance was so ingrained, that even if Apple had tried to sell a functional port of the Mac System 7 for PC hardware (something they had developed internally under a project called Star Trek), no existing software would work on those PC-Macs, because all Mac software was also inexorably tied to the 68k processor.

Without any software (no Illustrator, no PageMaker, no Quark, no Office), the PC-Mac system wouldn't be able to do anything but disappoint users. Porting all that software would have been far more difficult than just starting over from scratch. By the mid 1990’s, Apple realized its 1984 Mac OS needed a major overhaul anyway. Convincing developers to do the 20% of work necessary to get 80% there on PowerPC was difficult enough, and the new chip architecture seemed the the best foundation for building a new OS. The Pentium carried too much archaic legacy of its own.

Finding Processor Independent Nirvana

When Apple bought NeXT, they inherited technology built a generation after the first Macintosh. NeXT's engineers had introduced their system on the 68030, a much more powerful processor than the first Mac's 68000. That allowed NeXT to build an operating system with much more abstraction away from the underlying hardware, because they had the hardware power to make up for the additional software overhead.

The original goal of NeXT had been to build software libraries that would run on other operating systems, so the idea of not being tied to any specific hardware ran through NeXT's architectural design decisions on many levels. The end result was that NeXT's software was more portable, by design, than anything else like it, ever. Of course, there was also nothing else like it, either.

The rest of the world feared and loathed NeXT. IBM, HP, and Sun all made love-in partnerships with NeXT while at the same time working to destroy the company. Sun's Java Platform, and Apple, HP, and IBM's Taligent and CommonPoint initiatives all borrowed heavily against NeXT's technology in an effort to kill NeXT with a knockoff of their own stuff.

Microsoft jumped into the fray by announcing vaporous plans to deliver everything in NeXT real-soon-now in Windows; fifteen years later, they still haven't. Bill Gates famously said he would rather "piss on" NeXTSTEP than develop applications for it.

The collusion of the entire tech industry to bury NeXT ended when Apple snapped up the remains of the company in the last days of 1996. Suddenly Apple had the technology to move the Mac cross platform. They could put the Mac interface on top of NeXTSTEP and have it running on Intel PCs, PowerPC Macs, even on top of other operating systems, such as Windows NT and Solaris. Apple even hoped it would run on top of the existing Mac OS. This was Rhapsody: everything would just work, and work everywhere. Rhapsody failed miserably.

It turned out that despite Apple’s flashy, processor-agnostic new OS and framework technology, Mac users really wanted to run Mac applications, which wouldn't run very well stapled on top of Rhapsody in a compatibility environment. And PC users really were okay with running Windows, and not at all interested in dual booting another OS, unless possibly it were handed out for free. Apple's plans were ripped apart by market realities. Essentially, Apple had paid north of $377 million for StarTrek 2: yet another system that was portable, but stuck in place by non-portable third party software. Khaaaaaan!

Fortunately, even numbered Star Trek adventures are generally better than the odd numbered ones. Apple disassembled Rhapsody to develop a new plan: a new version of the Mac OS, which would only run on Apple hardware, but would offer the capacity to jump to another processor architecture in the future.

They applied lessons learned at NeXT to work harder at building third party support behind the strategy, and ended up dishing it out as Mac OS X, the obvious replacement for Mac OS 9. Mac users migrated to refined, "Carbonized" versions of their existing software, while Apple encouraged developers to start using the more advanced Cocoa frameworks they had acquired from NeXT.

Along the way, a few things happened: First, while the higher level APIs from the Mac (procedural Carbon libraries) and NeXT (object-oriented Cocoa frameworks) remained fairly distinct, their lower levels merged into Core Services. It incorporated support for Apple's existing technologies, which changed things dramatically from NeXT’s Foundation. Apple also rebuilt the display system using entirely new technology based on PDF imaging, and invested significant efforts into improving and modernizing the kernel, which similarly mixed in both Apple and NeXT technologies.

The result was that, while Mac OS X had far more in common with NeXTSTEP than System 7, it wasn't NeXT's old OS anymore. There was no Display PostScript, and there were no OpenStep compliant frameworks that could be easily ported to other operating systems. Apple had built a project to suit their own market's needs, not a scatter-gun, speculative project hoping to find a market, which had been the business plan NeXT had struggled with over its tortured existence as the red headed stepchild of the tech world.

Mac OS X was now an operating system for Mac computers, not a open specification, nor a cross-platform operating system for Windows PC users to consider buying, nor a development middleware designed to put a pretty face on Solaris, HP-UX, or Windows NT. These were the eggs Apple broke to make their OS X omelet.

Apple still suffered from the Star Trek problem: they could not move Mac OS X to more commonplace hardware and take advantage of the economies of scale in the PC industry until they'd weaned their Mac userbase off of the Blue Box (for System 7 / OS 9 compatibility), which was still tied to 68k emulation. They had to induce the migration of the majority of Mac software from Classic to Carbon. The secret card Apple held onto was developing Carbon libraries to be as hardware agnostic as the rest of NeXTSTEP had been.

After a surprise announcement just a year ago where Apple outlined a plan to introduce Intel based Macs within a year, and finish the transition by 2007, developers found out that Apple had been honing a new Universal strategy. It would allow both Cocoa and Carbon based applications to be rather easily compiled for either hardware platform. It wasn't going to be a transition from PowerPC to Intel, but rather the addition of a new hardware platform to Apple's development tools.

Developers are encouraged to ship their applications as Universal Binaries, not as Intel specific. This serves to create the smoothest transition to Intel possible, while also providing clear support for PowerPC Mac users throughout the normal lifetime of their machines. After the announcement, Apple released a new generation of PowerMac G5s with significant hardware design improvements, including the incorporation of PCI-Express and support for decent graphics cards.

Apple itself benefits from charting its migration on Universal rather than going Intel only. That's because the most invested software users will stabilize the market for PowerPC Macs (Microsoft Office and Adobe Creative Suite will not be Universal until next year), giving Apple the opportunity to release the first volley of Intel Macs to market segments that have less investment in PowerPC software but stand to gain the most relative improvement from the new Intel chips: the G4 based Powerbooks, iBooks, iMacs, and Mac minis.

While the initial announcement made it sound like the first Intel Macs would be released mid 2006, Apple ended up delivering all their consumer and portable lines in Intel versions well ahead of time.