This has an application, and it ties in to that ray tracing nonsense, and it ties into an after-hours programming project, but the basic idea is - what's the deal with boring old RGB values?
I wanted to get 20 discrete colors as separated as possible in colorspace, and this is what I came up with.
A less useful display of the same colors is here:
That looks an awful lot like the old Apple ][ low res graphics palette. But I'm a little concerned that the greens all look so similar. I know that human perception's a crazy thing, and RGB space isn't as useful as, say, YUV space for representing perceptual color. So, maybe I'll take another run at this in a bit.
But hey, I've just printed this out, and I'm going to cut it out and fold it to look at in more detail.
ETA: Ok, duh. I folded together the paper model of the picture above, and it's clear to me now one of the problems. I was worried that the greens were so similar, and it looked like the pinks were, too - there's a good reason, and I should have been able to see it in the unfolded image, above.
What I was trying to accomplish was spread the colors out as evenly as possible, but if you look at the two magenta-pink-ish dots in the center, they're practically on top of eachother, while the yellow dot is much further away from the white dot.
Ok, back to the drawing board.
You math geeks out there may be ahead of me on this one, but I just did some poking around and stumbled on this page:
And, in particular, there's a cube inscribed in a dodecahedron, which might be very close to what I need - it's a simple pattern, and I can do the mathematically simple projection, or I can tweak a single number (as opposed to three, which I had been attempting).
Also, all 8 of the vertices of the RGB cube coincide with vertices of the dodecahedron, which means the popular colors all show up for free.