You know, it's a little ironic, but the standard subdivisions of an inch (1/2, 1/4, etc.) or of fluid measure (8oz cup, 16oz pint, 32oz quart, 64oz half gallon, 128oz gallon), all of which are related by powers of 2, are a better match for binary arithmetic than powers of 10.

Sure, that's about the only place imperial measure works out better. But, I still find it a little ironic.

The powers of 10 in metric really don't mesh as nicely with computer arithmetic as you might like. If you start with meters, stick with meters until the end of the calculation. Converting to km then cm or whatever, will lose some precision due to the fact that ten isn't a power of 2.

Yeah, like nobody ever made the same mistake converting between cm^3 and m^3. 1 cm^3 = 1mL, and 1000 cm^3 is 1 L. 1000 cm = 10 m, but 10 m^3 is 1,000,000 L. But, folks in chem class made that mistake readily. Metric doesn't magically make the "forgot to cube the ratio" problem go away when dealing with volumes.

There's a more detailed article that the first one links to. In the more detailed article, it says:

Every year, I excavate about 2 to 3 cubic yards of material. I mine it from the walls during the winter, put it through the crusher, screen it, and then haul it out during a summer’s worth of Sundays.

So what we have is a unit conversion failure in the first article. 1 yard = 3 feet, but 1 cubic yard = 27 cubic feed. 3 cubic yards is 81 cubic feet, then.

So 7 * 81 = 567, which gives you a cube just over 8' on a side, as you suggest.

Not only that, but could it be that some of the pieces came from one place and others from others? For example, you could have one set of reactions near hydrothermal vents, filling the ocean with one set of building blocks, and another happening on land. A major land shift or oceanic event then mixes the two sets. Wash, rinse, repeat over a billion years. To me, the argument that it couldn't happen on land because what land there was was too unstable is more an argument that it could happen with pieces coming from both land and sea.

Well, it did do a lot for you. And the large RAM with flexible descriptor tables meant that in practice, you could avoid doing too many writes over to the VDP most of the time. And, the separate dedicated-RAM architecture does guarantee no cycle stealing, unlike, say, the VIC and VIC-II chips in the Commodore computers, or the need to wait for horiz/vert refresh to avoid "sparkles" like the old CGAs.

Let's face it, all these old computers were an exercise in tradeoffs.

I see the spot/crosshatch pattern on car windows when I wear polarized shades. I don't see Haldinger's brush, though, when I'm wearing no glasses. The pattern on car windows comes from the glass-tempering process.

I just had an image of Nigel Tufnel pointing to his left ear after such a test saying "This one goes to 11 . . . kHz."

Actually that's a good question: since you see UV light, could you use a UV flashlight to walk around in what appears to be almost complete darkness but you see just fine with the UV flashlight? I suppose that would be cool, not sure how useful that would be but interesting anyway.

Well, it'd be almost complete darkness, except for everything that fluoresces, which actually is quite a lot of things. Maybe he can get a job as plainclothes security at fun houses lit by black-light. Everyone else just sees teeth and the random glowing t-shirts and socks, whereas he sees everything else.

Narishma already said it here, but I was referring specifically to machines that used the TMS9918/28/29(A) VDP, often just referred to as "the VDP." So far as I know no system that uses the VDP was able to memory map and dynamically multiplex CPU and VDP access to the 4K or 16K of DRAM connected to it. And, that walled off access to the DRAM was a particular drawback for machines that used the VDP, which is why I pointed it out.