[Openicc] Helping with colord

[Chris Murphy]

On Mar 9, 2011, at 4:02 AM, Richard Hughes wrote:
> 
>> For the devices that report bogus EDID, the code that's building the profile needs to do some checking for rational primaries with some tolerance. A yellow x,y value for green isn't sensible. Negative values aren't sensible. Etc.
> 
> Yes, I suppose that's sensible, although I generally dislike heuristics.

In this case it is a kind of hack, but it's an immediate work around to the display profiling issue. If EDID isn't accessible, if the primaries are bogus, then the display profile is sRGB. Not much else can be done.


> 
>> And a major missing link persistently with camera JPEG is ignorance of EXIF color space.
> 
> Yup, GCM has a toy-API where you can give it a filename with path and
> it'll spit out a suitable icc profile filename. To assign the EXIF
> data to a profile, you just drag and drop the .jpg file into GCM and
> it creates a virtual device using the data from the EXIF tag. From
> there you can assign it a suitable profile. It's kinda a gimmick, but
> at least I know it's now possible.

The user has already assigned it a suitable profile. If it's in-camera JPEG, it's sRGB unless they've chosen Adobe RGB in the camera's interface. I would not necessarily make it difficult (let alone impossible) to reassign this color space, but as it is rare I would not invite such changes. If the user doesn't like the in-camera rendering, they're better off with a Raw workflow. Most of the reason why people aren't satisfied with their image has ZERO to do with the source space: #1 way above anything else is incorrect exposure, and #2 a close tie is having the light source set incorrectly, or to auto with a camera that doesn't estimate the scene light source very well. Both of these get baked into the JPEG and there's no way to unwind it.

Also, if there is an ICC present, it's supposed to take precedence over the EXIF color space tag. It is not necessarily the case the application that allows a user to change the assigned profile for an image should both embed the chosen profile, and then reencode that space into the EXIF color space tag. I think EXIF should be treated as read-only. The camera sets it, done.



> 
>> But still there needs to be a mechanism for competing proposals to be formally presented, debated and then something decided upon.
> 
> I think debate of design decisions is a totally good thing, but I'm
> not sure debating every little part of a perfect specification before
> implementation is a good use of anybodies time.

No, but I think the formalization of certain topics for debate and consensus allows a time frame for parties to be heard, and after that time frame a decision is made and we move on. If there's no formalization, then we get not just debating parts but endless debating. Also part of the formalization is it compels the proponents of each idea to be really susinct in their main points why their whatsit is a better way of resolving a problem, and why the other proposal is inferior. It establishes a kind of pier review, and something to reference.


> If Kai-Uwe turned
> around to me and said "hey, can you prefix all your display device
> ID's with 'display_' rather than 'xrandr_' and also include the serial
> number" then that's constructive. At the moment we're not having
> anywhere near that kind of dialog.

It's also a conversation that's limited to those who know what that means and why it matters. That sort of thing probably doesn't need to be formally debated. I'm only suggesting some level of formalization for certain topics that keep getting discussed over and over on the list without resolution. The informal method is not arriving at a decision. The formalization is a target for a future spec or standard, which for the opposition is also a bullseye for taking shots at, as it should be. And I think that this should be a brutal review process - of course the rules of debate preclude ad hominem attacks :-D But brutal does not mean endless. We also need a mechanism for concensus, and making decisions once that time has arrived. And then move forward again.

I think we're kinda stuck on an informalized analysis type of debate with lots of details. Formalizing debate means debate is about MAJOR points, not details, and there is a clear end to the debate, compelling a Next Step.


Chris Murphy