[Openicc] [Fwd: icm profiles in debian]

[Richard Hughes]

On 3 February 2013 10:13, Kai-Uwe Behrmann <ku.b at gmx.de> wrote:
> Looking at the colord package, all distributed profiles are automatically
> generated. In parts they use copyrighted names, which implies legal
> conflicts. That's the wrong way to do profile distribution.

Can you explain how using a copyrighted name implies a legal conflict?
I've asked about AdobeRGB before, and strictly colord should title the
profile "Compatible with Adobe RGB (1998)" which I'll do if anyone
from Adobe contacts me about the issue. I think it's quite obvious for
a "consumer" looking at the profile that it's not the official Adobe
profile.

> As well for normal users it is not clear, where those generated profiles
> come from.

Sure it is, colord adds metadata to that effect:

Object Path:   /org/freedesktop/ColorManager/profiles/icc_4532df47b7f58666d218ec8d842d422e
Owner:         root
Format:        ColorSpace..
Title:         sRGB
Qualifier:     RGB..
Type:          display-device
Colourspace:   rgb
Gamma Table:   No
System Wide:   Yes
Filename:      /usr/share/color/icc/colord/sRGB.icc
Profile ID:    icc-4532df47b7f58666d218ec8d842d422e
Metadata:      CMF_binary=cd-create-profile
Metadata:      CMF_version=0.1.29
Metadata:      CMF_product=colord
Metadata:      STANDARD_space=srgb
Metadata:      License=CC0
Metadata:      DATA_source=standard

> As soon as profiles are generated on the fly, like by colord during compile
> time, their profile ID will change with each second.

Sure, the ProfileID also changes if you add any metadata to the
profile, which means you can't take advantage of new features and
functionality. It's allegory to me is having a kernel module that has
to use a fixed kernel ABI -- it makes it impossible to add new
features without horrific hacks (like shipping the metadata in a
separate file).

> As the profile ID is
> the typical mechanism to detect source == output profile, that is for some
> systems a problem,

Right, but we can't pretend that's going to work when we use a sRGB
from any of the Linux profiles packages (e.g. openicc-whatever,
colord, etc) if we send the profile to anyone using OS-X, which has
it's own version. And we can't pretend it's going to work on Windows
either, which also ships an unofficial version of sRGB. If the user
wants this to work, then can manually download the official non-free
sRGB profile from the website and use that as the default sRGB
embedding profile.

I think it's much more sane to just check if the profile is identical
from the TRC and has the same matrix / LUT data, then it's the same
profile. If that's not acceptable, then we need to invent some kind of
metadata/private tag that we can compare (not something that changes
every time you change the profile metadata) and get Apple and
Microsoft to update their profiles and ship out updates to all
versions of their OSs. I just don't think that's ever going to work.

> (I am not on the colord-devel mailing list and would prefere to give
> suggestions here or in IRC.)

Sure, that's fine with me.

Richard.