embedded media norms and two separate text equivalents: alt text and descriptive text

Another proposal that should be considered alongside these other  
embedded media norms[1] is the question of whether the text equivalent  
for photographs should be treated the same as the other text  
equivalents that are indispensable for comprehending a document when  
the embedded media are unavailable to the user. To facilitate the  
greater expressiveness WCAG describes, a separate facility for  
descriptions of media would be useful. For example HTML4.01 has the  
longdesc attribute. HTML5 might use that attribute or add a new one  
like textdesc that gets away from the misnomer of long (since length  
is not really relevant to the purpose of the attribute, rather its the  
descriptiveness of the embedded media as opposed to its alt text media  
replacement).

Along those lines, one way to read WCAG is that such description can  
make use of a separate mechanism and is not always necessary to use  
the at text mechanism for such media description. In a certain sense  
using the same alt text mechanism for both alt text (media  
replacement) and media description degrades the user experience. By  
having two mechanisms and encouraging authors to make use of the  
appropriate mechanism for the appropriate text equivalent, users can  
grow to rely on finding media replacing alt text within the alt text  
mechanism and media description in the description mechanism.

So providing two separate mechanism for these two different types of  
text equivalent (alt text and descriptive text) is important (and  
lacking from the HTML5 current draft). Taking this one step further,  
these different text equivalents also have different levels of  
importance when evaluating the accessibility of a document. For  
example a document with thoroughly authored alt text is a very  
accessible document even if it is missing descriptive text. While a  
document with thoroughly authored descriptive text but missing alt  
text might have serious accessibility problems.

So if we consider this in light of the Flickr example, one reasonable  
reading of WCAG would suggest that the lone photograph on the page  
could reasonably have the alt text alt='', while the descriptive text  
might elsewhere describe the subjects and visual qualities of the  
photograph. However, due to the greater importance of the alt text (in  
this case merely null is sufficient), it would be reasonable to  
require alt text for photographs and merely recommend the descriptive  
text equivalent. In this way Flickr (and related authoring tools) are  
fully conforming to the HTML5 recommendation by merely including a  
null alt value. Flickr might also make it easy for authors to select  
photographs missing descriptive text and add such text equivalents.

So it may be a reasonable compromise — made possible by handling alt  
text and descriptive text through two separate mechanisms — to treat  
alt text as required and descriptive text as recommended for document  
conformance. This combined with the role attribute and the other noms  
I proposed earlier provides an effective solution to the problems and  
use cases raised surrounding text equivalents in HTML5.

So just to reiterate those norms with this proposed compromise:

  * the IMG element MUST include a role attribute with at least one  
suitable non-text media role keyword (decor, spacer, icon, logo,  
photo, albumphoto, portrait, chart, diagram, geomap, text, mathexpr,  
musicscore, livecam)[2]
  * the IMG element MUST include an alt attribute (though its value  
may be null in certain circumstances)
  * authors MUST include suitable alt text for each image embedded  
with the IMG element and authors SHOULD follow WCAG guidelines in  
composing suitable alt text
+* authors should include a suitable descriptive text equivalent  
referenced with the longdesc (or textdesc or whatever attribute name  
we choose; or even by providing user and authoring access to the  
immanent IPTC caption or another W3C specified property[3])
  * authoring tools SHOULD follow ATAG advice in assisting authors to  
compose suitable alt text and MAY automatically generate default alt  
text in cases where it is possible (e.g., the alt plain text of an  
image of richly styled text)
  * authoring tools MUST NOT add any text that is a placeholder for  
alt text (e.g., "this is an image")
  * authors MUST NOT add any text that is a placeholder for alt text  
(e.g., "this is an image")

Take care,
Rob

[1]: <http://lists.w3.org/Archives/Public/public-html/2008Aug/0581.html>
[2]: <http://esw.w3.org/topic/PF/XTech/HTML5/RoleAttribute#head-610fd0a42b1d8af2253378db31d9d28bd22988b9 
 >
[3]: <http://esw.w3.org/topic/HTML/UANormAndDOMForMediaPropeties>

Received on Friday, 22 August 2008 22:48:18 UTC