Macworld News Analysis: Setbacks won't stop Dual-boot Mac efforts

So new info (to me) in this article - EFI is not new technology, and has existed on the server side of the Windows world since the early 90's. Interesting. Then I found this series of posts by Daniel Eran, which explains it all better than I ever could have, and definitely sides with the Funke camp of Virtual Windows on MacIntel in the end. It's an excellent read, and if I were a teacher I'd make it required reading. The dude knows what he's talking about!

So EFI is a decade-old attempt by the PC industry to get away from the legacy hell of BIOS and move into the modern age of extensible firmware, much like Apple and Sun have been using on the PC side for years now. Apparently MS and Intel found it too prohibitive to migrate to EFI with their new systems, presumably because it would have broken all backward compatibility between new hardware and old OS software and caused all sorts of driver hell that would have alienated consumers and driven them toward alternate platforms, a.k.a. Macs or Linux. Apple naturally went with EFI over BIOS because of its extensibility and similarities to Open Firmware - they needed to keep their special boot modes and all the other firmware features that Mac users have come to expect and that OS X incorporates. They must have also accepted EFI with a sly wink, because it quite dependably blocks any easy attempts to boot XP. It will take, and is taking, a substantial amount of skilled, low-level work to emulate BIOS in EFI, which according to Eran shouldn't be too difficult for the right person, just time-consuming. Hence the delay in booting XP.

But wait, there's more. Once XP is booted on an Intel Mac, Eran notes that there's a fundamental incompatibility in the way BIOS motherboards and EFI motherboards handle the partition table, the crucial piece that tells the OS where to look for the OS and other data on the hard disk(s). BIOS uses the ancient MBR standard while EFI uses GPT. These standards are so incompatible that they cannot even exist on the same hard disk! Thus dual-booting between a BIOS-dependant OS like XP and an EFI OS like OS X will required separate hard drives. Yet another barrier to dual-boot.

Beyond that issue is the problem of device drivers, which would need to be custom-written for the MacIntel hardware devices. Eran notes that this problem could be solved by the release of VirtualPC for MacIntel, since the developers would be facing that same issue and would presumably have access to internal Apple hardware specs that third-party developers don't. Still, it's only a matter of time before the open source community reverse engineers all the hardware and writes drivers for the Apple devices.

Eran concludes, and I'm seeing the sense in this, that dual-booting isn't (and hasn't ever been) a good business model or a very usable configuration, and is especially unlikely given all the obstacles to "native" booting that MacIntel hardware presents to XP and 32-bit Vista. His conclusion is that virtualization is the wave of the future, and I'm forced to conclude, not unwillingly, that he's absolutely right. A virtualization layer between OS X and XP/Vista would eliminate the booting problems and the partition table problems by simply writing a low-level emulation layer between them. Beyond that, Windows apps that send instructions to the CPU/GPU would not need much, if any, translation at all, since the hardware is the same stuff that they would see in a pure Windows PC. Eran points out that MS was kind of caught with their pants down with respect to Virtual PC, however, since there has been nary a mention of when a MacIntel version will be released. It's a ground-up rewrite, but I presume that that rewrite will strip most of the emulation code out of the product, save for what I mentioned, and will thus run blazingly fast compared to the VPC of old. Throw in 3D game support and you've got a nice profit-maker for MS - MacIntel owners buying off-the-shelf copies of XP and Vista, maybe even in businesses that are mostly Mac but need a few PC-only apps, like Access or AutoCAD. Heck, I could use those in my Language Lab, where they use a Windows-only language program but otherwise don't do anything that couldn't be done on a Mac.

The only compatibility bummer remaining is that the Intel Macs don't support OS 9 "Classic" mode, which is a huge loss for the education community. I've already written them about this, begging them to reconsider and alter Rosetta to include Classic support. It's shaping up to be a major dissatisfaction point with technology teachers around the world, since the moribund education software market has abandoned many quality products in OS 9-land, never to be Carbonized, much less Universal Binarized. Sucks!