W3C home > Mailing lists > Public > www-style@w3.org > February 2012

Re: [css3-images] How does image-orientation interact with EXIF data?

From: Tab Atkins Jr. <jackalmage@gmail.com>
Date: Tue, 7 Feb 2012 01:11:08 -0800
Message-ID: <CAAWBYDB7NXzyCgWp6R+XF1Gty=N91WbM+F8EE-hK-hPha6EmPw@mail.gmail.com>
To: Simon Fraser <smfr@me.com>
Cc: "www-style@w3.org list" <www-style@w3.org>
On Tue, Feb 7, 2012 at 1:06 AM, Simon Fraser <smfr@me.com> wrote:
> On Feb 7, 2012, at 9:57 AM, Tab Atkins Jr. wrote:
>> On Tue, Feb 7, 2012 at 12:24 AM, Simon Fraser <smfr@me.com> wrote:
>>> CSS3 Images talks about images being encoded sideways, and how this can be corrected for presentation via the image-orientation property:
>>> <http://dev.w3.org/csswg/css3-images/#image-orientation>
>>>
>>> However, it doesn't describe how this interacts with EXIF data in the image. My understanding is that an image may be tagged with EXIF data that indicates that the camera was in landscape orientation when the picture was taken, but this does not affect how the image is encoded.
>>>
>>> I think the spec needs to talk about how image-orientation and EXIF rotation data interact.
>>
>> I don't think it's specified anywhere, but the web platform at this
>> point requires exif data to be ignored by default.  I'd be happy to
>> specify that somewhere in the draft.
>>
>> David had previously suggested giving image-orientation some controls
>> based on exif - in particular, the ability to flip an image (so you
>> can achieve the "unnatural" orientations) and an 'auto' value that
>> just uses the exif orientation directly.  This has also been requested
>> by some Chrome devs.  I was planning to do this in Images 4, but if
>> it's important I can add it now.  I'll raise an issue.
>
> So the problem that image-orientation is trying to solve is limited to that caused by images encoded differently, not those with EXIF data? If so, that seems odd, because orientation issues related to EXIF seem to be encountered much more frequently in the wild than those caused by encoding differences.

The spec inherited this property from existing print media stuff.  I
assume that such processors used the exif data themselves to figure
out what rotation to imply.  I don't condone this.

~TJ
Received on Tuesday, 7 February 2012 09:15:28 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 17:20:50 GMT