Given that RGB is well-known, I suspect their assumption was wrong, but I have nothing to back that up.

Not that it's very consistent. The color called "red" is #ff0000, which looks brighter than both #008800 and #0000ff IMO. And arguably, "blue" shouldn't be #0000ff, since that's just not a very pleasent color. I think I might've been tempted to go with something like, red = #cc0000, green = #008800, blue = #0055ff. Or maybe go all the way and specify the colors in some perceptual color space rather than using RGB. But oh well.

(The color #0000ff for "blue" is super over used IMO. Most terminal emulators default to it as well and it just makes blue text hard to read in most out-of-the-box configurations. To get a nice blue color, you need some green in there.)

If you used uncommon colors in your program on - for example - a Sparcstation 20, the palette would shift every time you moved the mouse in or out of your window. It's difficult to describe to someone that's never seen it, but it's unpleasant. No one mourns the death of pseudocolor.

#124356 might look different on one monitor or workstation compared to another.

Having colour names which are calibrated for the device makes a lot of sense. Assuming those colour names are actually calibrated, which as the article also mentions, so often wasn’t the case.

As an aside, this is a big problem in DTP where your display should match the page. However you obviously wouldn’t use colour names in that specific industry because you’re dealing with a vastly greater range of colours and shades.