Feedback on Responsive Images Extension

I wrote some thoughts over the weekend but it seems that the conversation has moved in a few different directions - submitting anyway... :)

This email is to provide feedback on the responsive images extension (http://dvcs.w3.org/hg/html-proposals/raw-file/tip/responsive-images/responsive-images.html#use-cases), specifically with regard to accessibility.

If the images are very similar in the difference picture's <source> options, then the spec is sufficient.  However, if the media queries are used to provide substantially different art direction (yes, I know that this is cautioned against, but that won't be sufficient to prevent issues), or if they are used alone or in conjunction with other scripting to determine what operating system or device the user is on, then it is easy to envision that the images might vary in very significant ways.  If, for example, the <picture> is part of an advertising block, the source for a user on iOS might be an advertisement for a new popular game for iPad/iPhone, but that the new popular game on Android or desktop windows could be very different.  Should authors use the capabilities to engage in browser or device sniffing, the spec doesn't seem to provide the flexibility to allow the authors to maintain compliance with accessibility standards.  I'm not sure that merely codifying that this is not what <picture> is supposed to be used for is sufficient.

To address this, I think the option of having alt on the <source> is needed.  The above example would be valid and appropriate for a set of like images, but something like the following is also needed:

<picture>
                <source alt="Angry Birds for iPad rated best game" media="(min-width: 18em)" srcset="med-1.jpg 1x, med-2.jpg 2x">
                <source alt="Voted best game for Android" srcset="small-1.jpg 1x, small-2.jpg 2x">
                <img alt="We have all your favorite games!" src="small-1.jpg">
</picture>

The processing rules would be that the alt value exposed to the accessibility API by the browser would be @alt on the picture (if present) + @alt on the utilized source (if present).  If neither existed then there would be a machine-detectable accessibility error.  The author would need to still monitor the quality of the alt and make sure that they choose the right approach, I believe that we  that we need this flexibility now and as the media queries become more sophisticated/capable these use-cases will become more prevalent.

There may be another solution to this problem (authors offering different images based on device/browser-sniffing and having incorrect text alternatives as a result) but that is the problem that I'm concerned that we'll see happening.

Thanks,
AWK

Andrew Kirkpatrick
Group Product Manager, Accessibility
Adobe Systems

akirkpat@adobe.com<mailto:akirkpatrick@adobe.com>
http://twitter.com/awkawk
http://blogs.adobe.com/accessibility