Turns out this is correct (search "correlated color temperature" in Google). Calculations are done in the 1960 Luv space by drawing the perpendicular to the Planckian curve through the u,v point having the coordinates of the distribution. Or one can draw the Planckian curve in x,y space with lines of constant correlated color temperature (not quite perpendicular in this space) and find the line closest to which the x,y coordinate of the spectrum lies (saves having to compute u and v). The locus of "D" white points is so close to the Planckian curve (p19 of the "White Paper") in xy space that you would probably incur little error by doing the comparison in xy space.

But sometimes Google / things you find on the Internet are wrong. :( Case in point: 6500K versus D65.

Anyways, I think we've really really gone off-topci from the original post. :D

Alas true but in this case Poynton confirms it (p223) and indeed running this algorithm against D65 returns 6504.2 K (published CCT 6504) and for illuminant C; 6775.4 K (published value 6774 K). I'm convinced.

Yes, a bit off topic but I needed to review this stuff for something else I'm doing related to the perceived color of beer and this discussion helped me do that. Thanks.