begin quoting John H. Robinson, IV as of Tue, May 24, 2005 at 03:44:30PM -0700: [snip] > This is easy to work around: Keep the supported hardware list small. > Apple already does this.
Sun does this, and it results (or seems to) in people getting irate about Solaris _not_ working on any-garden-variety-x86 box. People are already irate about MacOS X not working on 3rd-party PPC boards... [snip] > You can sell cheaper hardware (Apple is a hardware comany, right?) with > higher profit margins. Higher profit margins = higher profits if you can > keep the same number of units sold. Well, the premise falls down in that you're presuming the same number of units would be sold. It's as likely that units sold would *decrease*, as there's nothing to distinguish the box from anyone else's x86 box, aside from your OS not running on anything other than your hardware. Add in increased development costs, and you're facing lower profits, not higher. > > I don't think Apple makes that much money from the OS. I would bet that > > most of the income from the OS 'upgrade' fees goes to cover the advertising > > for that version of the OS. > > If the OS is a vehicle for the hardware: Stop advertising the OS. Use > your user group buzz to drive excitement. And why would that work? Microsoft showed us that advertising *works*, and works far better than user-group buzz and technical superiority. They demonstrated it so well that Intel *advertises* -- which kinda counts against that black-box theory as well -- to users who can't coherently describe what a CPU *does*. So why shouldn't Apple advertise? > How much advertising does > Linus do for the linux kernel? Not a lot, but then, a lot of people don't care about linux, and much of the time, you don't get a choice about upgrading. (Want the new glibc? Well, tough, you gotta upgrade your kernel.) > Give the OS away, or make it really dirt > cheap (like 25$ or so). Every so often, make the OS incompatible with > older systems, say every 7 years or so. Apple did offer a $25 upgrade, IIRC, but they bumped it up by $100 or so, possibly to offset the advertising, possibly to offest development of the OS, and possibly to make the upgrade seem more worthwhile. It does seem that people don't consider low-cost software to be worthwhile; they want to use (well, steal) the expensive stuff. And why make it incompatible except when you have to? Seems to me that you're advocating that Apple commit suicide. That's hardly good sense. > This way you force the users to a hardware upgrade path. You don't want > to be too greedy, otherwise people will flock to other OS's that wun on > that platform (Linux, *BSD) that will support older systems. Why force except when you have to? I resent Linux forcing me along the hardware upgrade path. (RedHat made my Alphas less and less usable over time with each successive upgrade. This is progress, I'm told.) [snip] > Unless you want to back up the word ``inferior'' with hard, fast numbers > then please leave the derisive language out. The x86 instruction set and architecture model sucks. It's nasty. Therefore, it's inferior to something less nasty, like the PPC. However, that's me. I know people (otherwise sane, caring, intelligent human beings) who _like_ the x86 instruction set. To the end-user? We're long past raw performance being an issue. What's needed is a CPU that supports the OS where the OS could use that support. My poor, slow, ancient 667-MHz Powerbook is as responsive as the 2.5-GHz WinXP box my coworker uses, except that large compiles take longer on my laptop. Since 90% of the time I'm _not_ compiling, I want the performance where it does me the most good. Whether that's in the OS or the CPU doesn't make that much difference, but I think that a good CPU helps the OS team to do a better job. At that level, it's not a black box anymore. > However, there are no price/performace comparisons. Only performance. Price isn't everything -- especially when it's so sensitive to the market. [snip] > Since the Intel is a CISC, it can do more per clock cycle than the RISC > PPC can. This makes the Dell a raw faster system by the numbers. A real > benchmark would provide better data. The article you quoted talked about this in detail. Remember, modern processors are running in starvation mode. Back in the mid-90s, my boss came back from COMDEX talking about a presentation he'd seen. They'd taken a 486/50 and loaded it up with a bunch of static RAM, and they'd taken a Pentium 90 with twice that in dynamic RAM... and the 486 toasted the pentium. By the mid-90s, it was obvious that the bottleneck wasn't the CPU, but the memory. So the CPU alone doesn't tell you very much at all about how much work can be done per clock cycle. Most of the time, the CPU is waiting on data anyway. > > People _like_ Apple because most of the time, It Just Works. > > This is because of control of the OS and the hardware. Solaris SPARC > Just Works, but is not aimed at the consumer, unlike WIntel and Apple. That's why my main box is a Sparc Blade 100. It Just Works. And works. And works. That's why I got it. Performance is nothing if it won't stay up long enough to do anything. (My secondary DNS is linux(debian)-on-sparc, but linux doesn't handle smp very well, so it crashes on large compiles. The reason I'm not running solaris on that box is because I want iptables.) > > Why upset that apple (ha!) cart? > > Why stagnate? Who's stagnating? Maybe the guy using an architecture based on a four-function calculator chip? -Stewart "Very sad that nobody found my pun amusing." Stremler
pgpfRvMjLzZjZ.pgp
Description: PGP signature
-- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
