467404 - ICC Profile for Philips 200WB looks bad with color management enabled

Status: RESOLVED DUPLICATE of bug 500404

(Core :: Graphics: Color Management, defect)

Description

[Emil Stenström]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; sv-SE; rv:1.9.1b1) Gecko/20081007 Firefox/3.1b1
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; sv-SE; rv:1.9.1b1) Gecko/20081007 Firefox/3.1b1

By request from bholley: When enabling color management, colors display incorrectly when using the default ICC file supplied by Philips for my screen (Philips 200WB). Might be a broken ICC file.

Reproducible: Always

Steps to Reproduce:
1. Set ICC profile to default color profile in Windows
2. Enable color management in about:config
3. Surf to http://friendlybit.com
Actual Results:  
See results in the attached screenshoot. The image in the back is with my Philips ICC file enabled, the front image is with the "sRGB Color Space Profile".

Comment 1

[Emil Stenström]

Attachment: Behind: With Philips ICC file, Front: With sRGB ICC file

Comment 2

[Emil Stenström]

Attachment: Color profile for Philips 200WB

Comment 3

[Bobby Holley (:bholley)]

Thanks for filing the bug.

I'm CCing Chris Murphy, our resident color profile expert.

Chris, the builds from bug 460629 don't detect this as a bad profile. Is there anything in the profile that looks fishy, or could it be a valid profile for some display (just probably not the reporter's)?

Comment 4

[Chris Murphy]

It's an embarrassingly sloppy profile, courtesy of Philips according to the copyright tag, and WGTI www.winglobal.com.tw according to the localized device manufacturer description tag.

The internal profile name, which is what will appear in application pop-up menus for ICC profiles, is "You always need to modify this tag".

It's not clear to me this was ever intended to be released into the wild. If it was found on a Philips web site, then they should be ashamed.

The profile has questionable green and blue primaries, considering the header purports the device model is "sRGB". The red primary matches up with sRGB OK. The primaries appear to be adapted, but the green and blue primaries are torqued such that the blue primary is more purple and the green primary is more cyan. It's probably wrong, but I don't know how the error would have been introduced. There's no work around except to shun the profile, and shame the source.

Expunge all bad profiles.

I'd like to take this opportunity in a public forum to point out how bad of an idea it was to not make the ICC profile display class file format simply colorimetrically based (measurement based) rather than requiring the profile vendor to do esoteric things like reading and following a spec containing bizarre concepts like chromatically adapting white points, and containing primaries within the realm of sanity. All of that mumbo jumbo could have been done by a CMM rather than asking clueless display vendors to figure it out.

Comment 5

[Bobby Holley (:bholley)]

Hrm - sounds like there's not much we can do in terms of general detection.

The best solution is probably to blacklist something about this profile, though I'm not sure what. Emil, where on the website did you get this? If we can find a large batch of philips profiles like this, maybe we can find some common characteristic (like that moronic internal profile name) and blacklist that. This would also require some sort of blacklisting framework for LCMS, which is beyond our current abilities but wouldn't be too hard to implement.

Comment 6

[L A Walsh]

I fail to see why this bug wasn't closed almost immediately as not being a bug in firefox.  As soon as it was determined that firefox wasn't doing anything 'wrong' -- it should have been treated the same as complaining about a web page for looking 'ugly' -- because it was designed that way.  I.e. it has nothing to do with anything related to Firefox and looks ugly in all rendering engines because it's poorly designed.  

Item's like this clutter up the database unnecessarily -- not that they should be deleted, comments about a bad profile may be useful to some extent -- but not as an open bug *against firefox*.

Comment 7

[Benoit Girard (:BenWa)]

This is either a valid profile and we should honor it if it exists, or this is a dupe of our faulty detection of bogus profiles.