[Openicc] GoSoC 2011: CPD and Color Management

[Chris Murphy]

On May 4, 2011, at 6:05 PM, Robert Krawitz wrote:

> On Wed, 4 May 2011 17:24:21 -0600, Chris Murphy wrote:
> 
>> Now that immediately raises the issue of sharing ICC profiles, which
>> all commercial products have varying license restrictions on how
>> they get shared. There is common practice which probably will not
>> really be punished under EULA's but in effect users are agreeing to
>> ICC profile sharing behavior that is problematic with respect to
>> those EULA's. In my view an ICC profile is not any different than a
>> Word, Excel, TIFF or JPEG document. I'm not restricted in any way
>> from sharing such files, although I am restricted from reverse
>> engineering their transforms - baked into those files. There is a
>> practical problem with a EULA that's as permissive as Microsoft's or
>> Adobe's (or countless others) when it comes to the "output file"
>> produced from an ICC profiling app, mostly because of the incentives
>> surrounding the baked in transform needing to be protected. If it's
>> not, one company (any company) could buy a copy of a profiling
>> application and then start selling profiles for $5 each. This would
>> cost Microsoft and Adobe very little, if any money, it might even
>> make them money as people want to read the contents of these
>> files. Whereas with the ICC profile, it's never a meaningful input
>> file for the application that created it.
> 
> That's flatly not acceptable: there needs to be a way to create
> profiles for Gutenprint that are unencumbered, or at worst (or perhaps
> best, given the point about being able to sell profiles), licensed
> GPLv2+, which most definitely allows reverse engineering the
> transforms.  Otherwise we (Gutenprint) will not be able to distribute
> profiles for our own package ourselves, and either users will have to
> find their own profiles, or distributors will have to cut deals with
> different outfits that lead to different output on different OS
> distributions with the same Gutenprint version.

Yes, small problem. I am unfamiliar with the Argyll license, although I do have it installed, so I don't know if it meets these requirements. Even if it does meet these requirements, and it's possible to populate its ICC profiles with print settings metadata, this is pretty much certainly not the case with any other application that creates ICC profiles. And I've thought as long as ten years ago this was a predictable outcome with the ICC profile format baking in transforms into the profile. It is actually at least as correct, if not more correct, to refer to profiles as "transformation profiles" rather than "device profiles" because while they can also describe device behavior, this is only true if they were produced correctly in the first place, and if you're smart enough to know how to unwind the transforms. They don't contain absolute measurement data only - it's highly manipulated data.

Getting away the notion that an ICC profile has IP attached to it in some way, we'd have to totally rethink the profile format toward absolute measurement data only profiles (that of course describe conditions, measurement devices used, etc) and leave even more effort up to the color management engine than at present. So it's shifts the problem around a bit, I think to a more appropriate location. But this doesn't really help matters today. We're stuck with what we've got, for the time being.

Chris Murphy