> It's not a good solution if you can use SATA spindles, but you'll require 6 times as many spindles and a bunch of SSDs, to get average acceptable performance, the cost will be high.

Why is it? Are not seek times inverse proportional to RPMs? So 7K SATA drives should give 30% less IOPS than 10K RPM drives. Does transport make any difference? I was under impression that SATA and SCSI variants are very alike nowdays. Is there difference in drives' firmwares? But then what is stopping manufacturers to put a better performing firmware on consumer devices? I really don't understand what factors can make 6x difference, could you please elaborate?

> You're still using a garbage collector? Do you watch that operate while gnawing on woolly mammoth bones or what? ARC is a far superior approach as it involves no overhead.

Ref counting is just a simple form of garbage collection that cannot deal with reference cycles. As far as I understand, Obj-C solution to the problem is "do not create cycles" and "use weak refs where appropriate". Actually I don't see how this can be definitely superior and the only right way as you state.

Why boring? There are a lot of fun things to do, and games are not really first in the line for everyone.

Reading nontechnical books, learning new programming languages and CS concepts, aimlessly rollerblading around, actually creating something needed by other human beings and communicating with clever people is what entertains me now the most. I used to play a lot in school and uni, first because of novelity, fast feedback and solvable challenges that later was satisfied with programming, then as another form of communication with close friends, then as some way of killing time and that didn't last long.

Generally there is a small ROM embedded in CPU that loads another bootloader from NAND, SD card, SPI Flash, etc. On Atmel ARM chips that bootloader must be small enough to fit into embedded SRAM. Than bootloader initializes SDRAM and fetches U-Boot into it. U-Boot in turn may initialize wider range of devices and then load Linux kernel.

All boot process is very SoC- and board-specific. Bootloaders must be compiled for selected CPU and board components, and Linux kernel should also have board description down to what types of chips are installed as autodetection is usually very limited.

Try one of Cyborg RAT mice. They are adjustable, and build quality is also very good. Not cheap though.

Well, single-core, 65nm, 600MHz OMAP3 in n900 consumes about 90mA@3.7V (~300mW) under full load, and about several mA when idle. 2W at idle and concept of smartphone are not really compatible. This CPU will discharge a typical 4W*h battery in 2 hours. And here It doesn't really matter if it is 10% or 100% faster then an ARM SoC.

Actually DVDs burn faster. 24x CD records at about 3.5MB/s, while 12x DVDs burns at 15MB/s. Both discs have approximately the same max rotation speed limited by media fragmentation, but DVDs have much higher density. So it's almost always more convinient to burn a CD image to a DVD, and if disk is rewriteble, than doubly so.