DNG profiles pretty much define the colour handling in the raw conversion process. So the format is documented and well defined.
ICC is intended to handle conversion between colour spaces, as far as I can see, but it doesn't specify the raw conversion process.
There have been thread here: http://www.luminous-landscape.com/forum/index.php?topic=82891.0
where amongst others an engineer working at Phase One (Esben HR) and who is said to maintain the processing pipeline clearly says that an amount of processing is done before applying an ICC profile.
As far as I understand his postings, the early stages of raw processing convert the image into a specific colour space, and the ICC profiles are applied to an image in that colour space:
"You will need to process the image in Capture One first, and select "Embedded" color profile. This will apply all the pre-processing steps, including non-linear curves, to the image, but the ICC profile will not be applied (it is embedded in the file). You want to create a color profile that replaces the embedded ICC, so the profiling tool needs to ignore the embedded ICC profile.
An image with an embedded profile will also contain the now-standard TIFF tag ProfileToneCurve (to go back to linear space). In theory, this should make it a much easier job for a profile creator, but they must obviously be aware of that fact. It could even require a different set of algorithms, depending on the exact way the software builds the profile.
It does require some fiddling to get this to work with tools I am aware of. X-Rite did a good job of making the process easy to apply for the supported Adobe products; it is not a simple thing to do from an engineering point of view. The quality of the resulting profiles are, I suppose, a subjective matter."
Although I have limited knowledge of the colour conversion process, that approach makes sense to. QPCard had some technical info indication that their ICC profile was not applicable to C1 and also stated that the reason was the need to go over TIFF. They say the next version which they have in beta test will support tiff reading and modifications to be able to support Capture One. This is entirely consistent with Esben's writing.
That just says it is a bit complex to generate ICC profiles for C1. It doesn't say it is impossible to use ICC profiles for conversion all the way, but I guess that you need to know what is the starting point.
Poster 'Vladimirovich' asked the question with regards to ICC profiles, and they do not strictly need a rendered TIFF, nor are they output referred profiles..., by common definitions. Andrew, who is otherwise known to provide good information (as a consultant should), somehow and for some reason only known to himself (judging from his posts here on LuLa) seems (!) to have decided to promote Adobe's DNG initiative, instead of keeping an open mind towards other market participants. It can't be just his personal situation that dictates that, or can it?
That's all fine though, if full disclosure is available, and truthful information is given, which 'appears' not to be the case. The responses have become more political, in the sense that one doesn't downright lie but just doesn't tell the complete (only selective) truth either. Hence the questions (to assert the veracity of his comments), not because people have watched the video or not. He should be pleased with the fact that people at least are trying to verify the value of his opinion, rather than simply label it as biased and irrelevant.
So, currently, there are two options on the table. Either Andrew didn't know that one can convert Raw data, before even demosaicing, to CGATS data for profiling (which would be surprising for a consultant of his reputation), or he deliberately left that aspect out of the supplied information (if so, why?) ...
I hope Andrew joins in the discussion, because I'd rather discuss matters with, than about, people. I value his information about color management, and would like to continue doing that.
So Andrew (if he's even reading this), what's up?