Since our original NTSC Compatible Color System was standardized in 1953, we've never had a color system that supports true Constant Luminance. Back in the HDTV 'argument' days, many, many engineers recommended that we switch to a system with Constant Luminance (CL). The NHK MUSE engineers attempted this with their "Quasi-Constant Luminance" system but, of course, that system is dead - the Japanese ClearVision EDTV system also uses a "quasi" CL encoding system.

Until recently, the argument against Constant Luminance was that televisions, which were all based on CRT, pretty much, would have to incorporate expensive gamma correction into the set - as a result, gamma correction has always been applied at the transmission end and failure of CL results, with only pastel colors being correct, and loss of true luminance resolution due to some of it being carried by the band-limited chroma channels... this also leads to things like "luminance notches" at the edges of stuff like credits of certain colors as well as incorrect color reproduction for anything saturated.

Nowdays, gamma correction is in pretty much every digital display and CRT's are a thing of the past, so why not switch to a CL system? A system with constant luminance 'looks' like it has higher chroma bandwidth, even when it doesn't, as compared to a non-constant luma system.

There's plans for deep color, etc... but nothing for Constant Luma that would really increase color fidelity and picture quality.

Anyone know the reasons behind this nowdays?

By constant luminance, do you mean a linear gamma (1.0)?

That makes signal processing easier, but you need a lot more bits to get the same level of accuracy. Human vision has a gamma response curve very similar to the 2.2 gamma used by CRT and digital media, so even without CRTs, we'd want to use 2.2 gamma for 8-bit and 10-bit codecs. With linear gamma, you wind up having vastly more precision in the brighter part of the region, wasting a lot of bits.

To use linear gamma, I really like to be in 32-bit floating point. I use linear 32-bit float processing in After Effects all the time, but that's just a processing thing; I never store it in a file.

So, broadly, if the goal is to reduce banding, just going to 10-bit with today's 2.2 gamma is the best, cheapest solution.

And that said, appropriate dithering can reduce nearly all 8-bit banding you might see today. The bulk of banding people see on HDM titles was presen in the source anyway.

No - Here's a link that explains it much better than I ever could.
http://www.poynton.com/PDFs/Mag_of_nonconst_luminance.pdf

By constant luminance, do you mean a linear gamma (1.0)?

That makes signal processing easier, but you need a lot more bits to get the same level of accuracy. Human vision has a gamma response curve very similar to the 2.2 gamma used by CRT and digital media, so even without CRTs, we'd want to use 2.2 gamma for 8-bit and 10-bit codecs. With linear gamma, you wind up having vastly more precision in the brighter part of the region, wasting a lot of bits.

To use linear gamma, I really like to be in 32-bit floating point. I use linear 32-bit float processing in After Effects all the time, but that's just a processing thing; I never store it in a file.

So, broadly, if the goal is to reduce banding, just going to 10-bit with today's 2.2 gamma is the best, cheapest solution.

And that said, appropriate dithering can reduce nearly all 8-bit banding you might see today. The bulk of banding people see on HDM titles was presen in the source anyway.

Totally agree.