@Ofnuts, okay, that's what I meant. :^) Linear light is what I want.
@rich2005, I've been misleading myself by using blind-to-all-but-8-bit "Colors->Info->ColorcubeAnalysis" to "check" how many graylevels are in my 16-bit gradient and convincing myself it wasn't really 16-bit.
But... I don't get a lot of reassurance when I use the "Color Picker" to set a foreground/background color; it seems to report something between 0.0 and 100.0 (in my case with 1 dec. pt. precision, implying 1,000 colors).
Now what I haven't checked... is IF I can set a colors like, say, 012345 and FEDCBA hex as foreground and background colors, run a linear CIE gradient and see what kinds of values Color Picker shows near the endpoints.
And so I checked. Both 012345 and FEDCBA appear are shades of blue when selected as foreground and background colors, but using them with the Gradient tool (blend:CIE shape:linear) produces something that has only 256-levels of gray: sampling anywhere in the result produces a color in RRGGBB format with RR=GG=BB--256 levels.
So while I can then access ANY "color" from 000000 to FFFFFF, they do not map to 65536 (actually) levels of gray.
AND NOW, it dawns on me like a falling load of bricks what Ofnuts has been repeating: 65536 "unique" colors when converted to still only map to 256 unique levels of gray.
I found
this example on stackexchange which I would not have gotten from the
wikipedia article on CIELAB/L*a*b color space.
Attachment:
CIELAB_50_0_0.png [ 28.7 KiB | Viewed 665 times ]
(I'm almost done. There remains one question.)