Speaking to you and Graystar, I'm not disagreeing with you that the accepted way is the 'right' way to do things 99.9% of the time. It makes sense; when we move from one color space to another, we want to preserve the appearance as much as possible most of the time. That doesn't mean that joofa doesn't have a point and that considering it might lead to an interesting discussion of when it might be a good idea NOT to use chromatic adaptation during color space conversions. I don't see this as a black and white scenario (pardon the pun).
And that's how I read Bruce's comment. 99% of the time, AdobeRGB is fully contained in ProPhoto RGB. However, if you do a certain thing which has some large consequences (R=G=B is not neutral), it's not. That certain thing is the example that joofa pointed out.
It's probably worth just putting this image out there for people to judge for themselves—if you are interested in the discussion, but you are not as you say 'in the industry' a visual sample might be informative. I'm just throwing it out in case a concrete example is useful.
The grey square on the right is roughly (given the vagaries of the internet) neutral grey defined with equal RGB values in AdobeRGB 1998. It should look achromatic. The square on the right is a conversion to ProPhotoRGB with no chromatic adaptation. You could achieve both these colors
appearances with the same XYZ coordinates by changing the viewing conditions. But the question, the question this whole thread boils down to is: when you look at these patches under the same viewing conditions, like you are right now, are they the same color? In other words, do they look like the same color.
It was really hard to parse Joofa's meaning, but from what I could gather he wanted to build a case that if we call these two patches the same color, if we start from premise that is questionable at best, then the colors in the AdobeRGB space can be outside the ProPhotoRGB space. That's when all the analogies like "If my aunt had balls she's be my uncle came around." (If that's all I get out of this thread, it's worth it.) But they're not the same color unless you are willing to let XYZ coordinates override you eyes and brain and define color rather than model color. To do that is a misplaced concreteness (
http://en.wikipedia.org/wiki/Reification_(fallacy) ).
There are times like when proofing one paper/device on a different paper device that you want to ignore white points and create color casts. It lets you do things like proof a dull paper on a bright one. But if you spend anytime working with digital imaging, you won't buy that argument in this case. The profiles in question represent
working spaces. Which means they are abstract environments in which to work. The number Bruce Lindbloom gave: 99.9% is being generous. I can think of NO time when converting between working profiles that you don't want neutral to stay neutral. None. If it happens you assume something is broken.
So when Joofa shows us that, under a condition that should never be the case, a well-known fact is not true, you must be able to understand that he will be met with some pretty stiff resistance. When he is exceptionally cagey in his responses to questions and is unable to give a clear explanation of what he's doing, but insists that he is right and intimates that some well-respected members of this community don't know what they are doing, he becomes a troll.
