Re: acceptable fallbacks [was: Re: Is longdesc a good solution? ...]

per the last item, I don't have a sound card rendering me both deaf and 
blind so Please make the appropriate substitution in your reference?

----- Original Message ----- 
From: "Henri Sivonen" <hsivonen@iki.fi>
To: "Jim Jewett" <jimjjewett@gmail.com>
Cc: "HTML WG" <public-html@w3.org>; <wai-xtech@w3.org>
Sent: Wednesday, September 10, 2008 3:37 AM
Subject: Re: acceptable fallbacks [was: Re: Is longdesc a good solution? 
...]



On Sep 9, 2008, at 22:34, Jim Jewett wrote:

> Henri Sivonen asked:
>
>> Would this
>
>   <video src=movie.ogg>Please upgrade to a browser
>   that supports HTML5 video.</video><p>
>   <a href=transcript.html>Annotated transcript</a></p>
>
>> be a "hack"?
>
> As a usability issue, please *at least* reverse the order.  Links
> following a "Please upgrade your browser" messages are almost always
> to downloads of the suggested browser or plugin.  (Alas, that
> conflicts with the standard desired visual layout unless CSS is used
> to reorder.)

I put the "Please upgrade" stuff in there to emphasize that the
content of the <video> element is not accessibility fallback but
legacy browser fallback. Per spec, the text about upgrading is not
rendered in <video>-supporting browser regardless of user's disability.

>> Is a semantic association between the <video> element
>> and the  transcript necessary if the link is very near the video
>> in the document reading order?
>
> While the above is adequate, it is still a hack.  It relies on people
> continuing to leave them together, and to recognize that they are
> together, and to not have stronger prior assumptions about nearness.
> It is still better than most existing sites; the question is whether
> it would be worth an explicit annotation.
>
>   <details id=mov1><legend>
>     <a href=transcript.html>Annotated transcript of XYZ</a>
>     </legend> ... </details>
>   <video src=movie.ogg fallback=#mov1>
>      Please read the transcript or upgrade to a browser that supports
> HTML5 video.</video>

That doesn't look like proper use of details when the legend is also a
link instead of the transcript being in the content of <details>.
Also, experience with usemap suggests against "#mov1" and in favor of
"mov1" ID ref.

This would be better:

   <video src=movie.ogg transcript='transcript'>Please upgrade to a
browser that supports HTML5 video or <a href=movie.ogg>download the
video file</a>.</video><p>
   <a id='transcript' href=transcript.html>Annotated transcript</a></p>

With a rules that
  * The transcript attribute is an ID reference.
  * If it points to a <a> element, the transcript is the resource
pointed to by the href attribute of that element.
  * Else, the transcript is the content of the element pointed to.

> (For the moment, I am explicitly not taking a position on whether that
> @fallback should be @alt, @longdesc, link@rel=alternate, etc.)
>
> I'm almost inclined to say "no", because the final <source> element
> could do the same job, if properly defined.
>
>   <video src=movie.ogg>
>     <source src=movie.ogv type="video/ogg">
>     <source src=#mov1>
>     Please upgrade to a browser  that supports HTML5 video.</video>
>   </video>
>   <p><a id=mov1 href=transcript.html>Annotated transcript</a></p>

This assumes that the fallback worked on user capability as opposed to
codec compatibility.

For images, falling back to text makes sense when either the user
turns off image loading so that the CSS formatter can render text in
place of the image or the user interacts with the page through AT that
renders to different media (audio or braille) without interfering with
the CSS formatter.

Unless the user is both blind and deaf, it doesn't make sense to turn
off video loading categorically, since a person who can't see the
video track of speech-dominated videos can listen to the sound track
and a deaf person may still watch videos where the visual content
dominates or where captions are available. For this reason,
transcripts shouldn't be modeled as automatic video fallback and
should instead be modeled as content that any user can navigate to.

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/

Received on Wednesday, 10 September 2008 12:06:46 UTC