Sure, it has been rumored for years. Sure, any technology observer even slightly familiar with Apple knew that Mac OS X had been run in house on Intel. But, Apple parting ways with the processor it has spent all of these years promoting? If Apple was a few millennia older, without doubt this would have been prophesied as a sign of the apocalypse. The real apocalypse may not be here yet, but the computing world has just seen one of the biggest earthshaking announcements in years. Now Apple faces one of the hardest projects ever put forward for a computer company in its position: keeping backward compatibility.
So my predictions were off the mark. I said I doubted that Apple would switch, I doubted that they could do fat binaries and I doubted that they could pull off a decent emulation layer. Could it be that I was wrong on all three counts? The first, most definitely, was wrong, as we now know that Apple is making the switch to Intel. The second was wrong as well; the name may be “universal binaries,” but the fact remains that what we see is essentially the fat binary concept returned from the dead. So what about the third?
This is what will prove interesting. Jobs, according to observers at WWDC, demoed Microsoft Office 2004 and Adobe CS running on an Intel-based Mac — unmodified — thanks to what Apple is calling Rosetta. MacNN reports that it appeared to do so “without significant speed decreases.” Apple, for its part, concedes that PowerPC programs will run slower on Intel systems. Of course, questions still remain in the air: how much had Rosetta been tuned for that specific demo of two applications that are already being pledged to come in x86 forms, and assuming it was tuned, how much compatibility can Apple achieve in the next year? Perhaps even more importantly, how insignificant, or significant, are the speed decreases?
According to documentation released by Apple, what Rosetta will not do, is provide compatibility for applications that directly call on AltiVec (as opposed indirectly using it through Mac OS X 10.3 or 4's acceleration libraries) or programs that are compiled specifically for processors older or newer than the G3. This means G4 and G5 optimized code will not run using Rosetta. Perhaps even more importantly, the documentation also states that Rosetta will only work on Mac OS X binaries, not Classic application binaries.
These are the kind of questions that are not yet clear, but we anxiously wait to get a better view of. If — and this is a big if — Apple can pull off making Rosetta provide near native speeds, perhaps Apple can gain by switching to the next generation dual core Pentiums. Where this puts 64-bit computing is a good question, but Intel may have new 64-bit improvements up its sleeves. While the likelihood that Intel would do major changes just for Apple seems slim, it is certainly conceivable that Intel may focus some extra work on items of interest to the next generation of Macs.
Other questions arise, such how much larger universal binaries will be as compared to standard binaries, what kind of performance changes we can expect running such binaries on PowerPC systems, and perhaps most importantly for short-term adoption plans, how long Apple plans to release PowerPC versions of its future OS updates.
This is an interesting time for Apple; there is no doubt about that. Assuming the company can pull off another unprecedented switch, it will again prove that, like the cats that have provided codenames for Mac OS X releases, it has many, many lives.