[Openicc] ISO - Open vs. Free (was Linux CM ideology)

[Chris Murphy]

On Feb 12, 2011, at 12:21 PM, Chris Lilley wrote:

> On Saturday, February 12, 2011, 1:44:14 AM, Chris wrote:
> 
> CM> I will agree with Leonard's comments. I commented a number of
> CM> times on CSS2 and then CSS3 long time ago on the very antiquated
> CM> section on gamma for various operating systems, and also on the
> CM> proposed tags for CSS3 that would have allowed tagging without
> CM> embedding, and how that needed to be cleaned up a bit. Nothing
> CM> happened. No one changed anything.
> 
> That is incorrect; the section on gamma for various operating systems was removed from the spec, due to your (and other's) comments.

Well it's nice to wake up on a Monday morning with egg on my face, eating crow for breakfast, when something like this has actually been fixed. While the W3C website is like wading through quicksand, after 20 minutes of looking around I see in the current 2.1 working draft (which is also last call status?) the gamma section is in green, and considered a non-normative note. I guess that means removed.

> CM> And then finally after
> CM> some time I mentioned it all again for CSS3 and what I was told
> CM> was basically it was too late. They were pulling all of the color
> CM> tags out of CSS3 because no browsers had implemented support for
> CM> them, and yet they weren't removing the b.s. gamma section even
> CM> though that has never been implemented by browsers either, and is
> CM> also factually untrue, and not good advice anyway.
> 
> As mentioned, the gamma section was removed.
> 
> Yes, the part of overriding the rendering intent of embedded images in profiles was removed. It wasn't implemented and probably wasn't a good idea, either.

Well it looks like the dropped features for CSS3 include the color-profile and rendering-intent property. I have less of a problem with the rendering-intent property than I do the color-profile property overriding embedded profiles. If the wrong profile is embedded, overriding is not a good way to specify the correct profile. The embedded profile in the image needs to be changed. Otherwise we end up with a PNG scenario. And when CSS gets used for other purposes than in a web browser, and a document is converted to PDF, what's supposed to happen? It would have to strip the embedded profile from the object, and embed the override, because PDF won't let you have two profiles per object. I think that's asking for a lot of 2nd guessing work by UAs and a lot of troubleshooting....messy.

And as notes point out, names aren't reliable, and mentioned MD5 which isn't specified for v2 profiles (99% of the profiles on the planet), and isn't required for v4 profiles (i.e. MD5 field can be zero). So the whole 2nd guessing concept is highly questionable at best. So I'm glad it got dropped, but I'm not please auto and <url> have been victims as well. Baby with the bath water.

The one thing I don't like about <url> is how it can override embedded profiles. Embedded profiles are sacrosanct. If they're wrong, they need to be removed or changed. We can't have UAs second guessing embedded profiles.



> 
> CM> So what's old, wrong, and not implemented is what's in CSS3. What
> CM> could have been useful with modifications, went no where.
> 
> Both statements incorrect.
> 
> The gamma stuff was pulled from CSS *2.1*, not 3.
> 
> The stuff that could be useful with modifications was modified and did go somewhere.

OK I'm not finding anything other than whole sale removal in the case of CSS3. And not removal, but relegation to non-normative status for CSS2.1.

> CM> So from my perspective, this expert's advice for the W3C totally
> CM> fell on deaf ears. 
> 
> I'm sorry if you didn't get good feedback from your comments. But they did have an effect, even if belatedly (I only got involved with CSS again a couple of years ago).

Apology not required. It's not your fault the W3C is like back to back weeks of different cuts of Dune playing over and over again.


> 
> CM> And considering it takes epochs for the W3C to
> CM> get things done, it might be 20 years before there's another
> CM> opportunity for a CSS3.5 or 4 the properly accounts for color.
> 
> Since 'getting things done' implies having a test suite and demonstrating that at least two implementations pass each test, yes, it can take a while to get to the final standard.
> 
> Chris, has your earlier experience with W3C soured you to commenting again, or should I invite you to comment on the spec where the actual colour management stuff is?

Both. I definitely prefer commenting though.


Chris Murphy