Stardate 20030511.1737

(On Screen): In my post yesterday I mentioned that it wasn't possible for Apple to switch to selling machines based on one of the X86 family processors. I wrote about that problem last year and explained the reason why, but it was buried in a much larger article, so I wanted to visit that again and explain more fully why it won't happen.

Partly that's because of some of the comments I received. A point made by several was that OSX runs on the Mach kernel and is based (at least a bit) on OpenBSD Unix, and that both of those are highly portable to the x86. All of that is true, though it should be pointed out that the largest part of OSX itself is proprietary Apple code such as Aqua and the rest of the GUI. However, I think the presumption in these comments was that the kernel and low level OS stuff are the parts most likely to involve a lot of assembly code and other machine-specific things which would be hard to port, and the assumption that the higher level code (such as Aqua) is written in some sort of high level language and could be moved to a different architecture by using a different compiler.

I'm quite willing to grant that Apple has the ability to make OSX run on an X86 platform. I would not be at all surprised to learn that they have it running on such a platform now in their labs. I would not be surprised to learn that it was reliable and had excellent performance.

I am quite willing to grant that Apple could manufacture a Mac based entirely on commodity hardware from the X86 world. Apple could produce a Mac which used the P4, or the Itanium, or the Opteron, or the Athlon64 come the day. Apple could produce a Mac which used two Xeons, or two Opterons, or two Itaniums, which fully utilized both processors. They could make it take advantage of hyperthreading in the P4 and Xeon.

All of that is possible. It might even be easy. And all of it misses the point.

Bill sent me a pointer to an article by John Dvorak written in April in which Dvorak says not only that he thinks it could happen, but that it will happen and that because of it Apple might actually begin to take back substantial market share. And Dvorak too misses the point.

It's possible to describe things which would work and would be useful, and show how they'd work once they were in place. I've run into this kind of thing before. Back in the day, when I was a cub engineer and worked at Tektronix, I knew a guy named Keith Lofstrom who had an idea for what he called the "launch loop". It's an interesting concept, to be sure. At the time I spent some time looking at Keith's idea and tried to point out such flaws as I saw in it, and by far the most important was that there were a number of potential ways in which the system could fail, and every failure mode I came up with was catastrophic and destroyed the system. (Poul Anderson Frederik Pohl actually put a "Lofstrom Loop" into one of the Heechee books, and had it fail catastrophically.)

It only really occurred to me later that the real problem with the system was that it wasn't possible to build it. You could describe how it would work once it was in place, but there was no obvious way to make it start. In practice it turned out that the only way to build a launch loop was if you had a lot of other technology which does not now exist, and if you had that technology you wouldn't need the launch loop any longer.

The current ideas about the "space elevator" suffer from the same problem. There's a research institute (I'm tempted to put scare quotes around that but will resist the urge) which had a bunch of research papers and briefings online which I looked at after they tried to interest some Australian city in investing in the idea, and all of the papers concentrated on analysis of the system in its final form, looking at things like the stress which would be placed on the ribbon itself to see if materials exist which are strong enough.

I have serious doubts about the entire concept even in operation, which I may write about some time. But the biggest problem, once you've found the funding, and have all those hundreds of miles of Kevlar ribbon sitting on the ground in spools, and have built the anchorage, and have the elevator ready to go, then how do you actually put the ribbon into place? Those research papers tended to wave their hands and skip over that part, because it's an extremely tough problem and may require other kinds of technology which don't exist and which, if they did exist, would obviate the need for the elevator itself. Without such things as rockets capable of generating high thrust for weeks at a time without refueling, the problem becomes extremely difficult even if the anchorage is on the equator. (Putting the anchorage in Australia only made the problem worse.)

That's the same problem that we face with respect to an x86 Mac running OSX. I will grant that Apple's engineers are capable of producing such a thing. I'm even willing to grant that they've already done so.

Equally, looking ten years out I can understand how such a product, once established, would be viable and competitive. The problem is that I don't see any way to get from here to there. Even Dvorak waves his hands and skips the biggest problem:

Late 2004 or sometime in 2005: Just as Apple's ne