It's among the most famous skunkworks projects in personal computer history. Marklar was a secret Apple development program to put Mac OS X on the Intel x86 microprocessor -- or in some sense put it back on Intel, as the underlying NeXT OpenStep OS did work on x86 chips. Ten years ago this August, the first whispered rumors of Marklar emerged.
Considered a fallback plan in case the PowerPC architecture couldn't keep up its performance curve, the Marklar team worked off campus to prevent further leaks while maintaining a feature-identical version of OS X on Intel for years -- all the way through Jaguar and Panther. Even though some of us didn't believe it possible, Apple eventually demoed the Intel version of the Mac's OS in 2005 and released it in 2006 as Mac OS X Tiger 10.4.4.
The secret history of Marklar is the stuff of legend, but the story of the original idea for an Intel version of OS X hasn't ever been told, until now. In a Quora thread about how Apple is so effective at keeping products and projects under wraps, Kim Scheinberg explains how her husband John, a veteran Apple engineer, wanted to find a project that would allow him to telecommute from the East Coast -- the couple had a one-year-old son, and they wanted to be closer to family.
John (or JK), who had actually risen to the rank of Director within Apple before "self-demoting" to Engineer so he could do more coding and less managing, sent an email to his boss in June of 2000 suggesting that he be given leeway to work on an Intel build of OS X. Permission in hand, he proceeded to work on the Intel build for eighteen months and had it running on a total of six PCs before his boss asked to check in on his labors.
John showed his boss Joe the bootable, functional OS X on Intel -- that's where things got interesting. As Scheinberg writes:
Joe pauses, silent for a moment, then says, "I'll be right back."He comes back a few minutes later with [former SVP of software] Bertrand Serlet.[Our son] and I were in the office when this happened because I was picking JK up from work. Bertrand walks in, watches the PC boot up, and says to JK, "How long would it take you to get this running on a (Sony) Vaio?" JK replies, "Not long" and Bertrand says, "Two weeks? Three?"JK said more like two *hours*. Three hours, tops.
Serlet had them buy a Vaio, and sure enough it was running JK's build of Mac OS X within hours. That was the start of the skunkworks project; the development effort kicked off in earnest after that demo in December of 2001, and the team was moved off Apple's campus immediately for secrecy's sake. TUAW's Michael Grothaus, who was in Apple's sales group during the latter part of the Marklar development curve, notes that the project was spoken of in half-believed rumors and mystery even within Apple.
When you read the full story, take a moment to consider what the Mac might be like today if not for John and Kim's family values. John's long retired from Apple, so Kim felt comfortable finally sharing the tale -- and for that, we thank her.
If you're wondering how Marklar got named Marklar, Scheinberg has an answer for that too: she named it. When JK asked her for a name for his third PC, she grasped at random for Marklar, which became internal shorthand for the OS development project itself. By the time he left Apple, most of JK's machines had South Park names. [Corrected]
Update: Red Sweater's Daniel Jalkut, who was responsible for Carbon API support in Apple's OS group during the late '90s and early 2000s, sent a note regarding the overall awareness of Intel capability on the engineering side:
As unexpected as Marklar and i386 support was to the general public, many engineers at Apple were not surprised, per se, to have their code running on i386. Since Cocoa ran on i386, as a matter of good hygiene many folks left i386 as a compiler target at least. We weren't actively running or testing i386, but the discipline kept our code more flexible. This meant that when the choice to adopt i386 became clear, it was not as massive an undertaking as it would have been if no thought had been given to cross-platform compatibility over the preceding years.
Thanks to Kim Scheinberg for her corrections to this post.