Re: ALT, <video>, and a different point of view

On Thu, 6 Sep 2007, John Foliot wrote:
>
> Ian, I don't need or want explanations, I need and want proof (exactly 
> the same thing you demand of me and others).

I think you may have misunderstood what I've been asking for. I merely ask 
that data be made available where available, that research be completed 
where possible (e.g. Joshue's very useful videos). I'm not sure what 
"proof" would even mean in this context; this isn't a mathematical 
science, after all. It's just language design.


> 1) you made the statement: "Anything they add (filename, tags, title, 
> comments, EXIF data) would actually be _worse_ than simply not having 
> any alt text", to which I asked how you came to this: please point out 
> the specification, user-study, quoted authority or independently 
> verifiable source to support this claim.  Where is the data?

My claim is based on the concept that a user only needs to hear 
information once, not twice.

For example, this:

   <figure>
    <legend> My house. </legend>
    <img alt="My house." src="house.png">
   </figure>

...is an example of useless alternative text as it just duplicates text 
that the user already has. Why would the user rather hear "My house: my 
house" rather than just "My house"? It doesn't improve the user's 
experience.

On the other hand, this:

   <figure>
    <legend> My house </legend>
    <img src="house.png">
   </figure>

...is mildly better (it can just be read as "My house: image", or "My 
house: image house dot ping").

Of course, even better would be:

   <figure>
    <legend> My house </legend>
    <img src="house.png" alt="I live in a small blue house. The front 
    door is green, as is the lawn in front of it. There are two floors,
    the upstairs windows have geraniums growing in baskets.">
   </figure>

...but the whole point of this discussion is the cases where the system 
doesn't have any alternative text available. (Most people uploading photos 
to Flickr don't seem to even bother with titles on all their images, let 
alone useful textual equivalents.)


> 2) you talk of "advanced heuristics designed by accessibility experts" - 
> what heuristics?

There is a large range of heuristics from the naiive (reading out the 
filename, as is done by many UAs today, or saying "image with title" and 
then reading out the title, or describing the image's dimensions), to the 
bleeding edge (e.g. analysing the image in the context of other images on 
the site and establishing likely three-dimensional relationships between 
images using techniques like PhotoSynth's [1]) all the way to heuristics 
that are yet to be developed (such as examining the image and describing 
its characteristics, or recognising features in the image [2]).

[1] http://en.wikipedia.org/wiki/Microsoft_Live_Labs_Photosynth
[2] http://labs.live.com/photosynth/whatis/smartphotos.html


> Which accessibility experts?

Any. I'm assuming that accessibility research is not a field that has 
reached its peak. I sure hope I'm right, and that the future will allow 
blind users to use computers to have images usefully described to them, 
and deaf users to have audio tracks automatically transcribed for them, 
and so on.


> 3) you state: "Specifying alt="" on a critical image is actively 
> *harmful* to accessibility. That's exactly what we *don't* want to 
> encourage!" to which, once again I request that you point to the 
> definitive source of this claim.

Given that the image is critical, and that critical images convey 
information that the user may want, and given that alt="" is indicative 
that the image is irrelevant (e.g. decorative), and that irrelevant images 
are intentionally ignored by the user agent so as to not distract from the 
actual content of the image, labelling an image that is critical as 
decorative would make it much harder for a user to determine what image 
content is actually critical. This would be harmful.

To put it another way, this:

   <p><img src="separator.png" alt=""></p>
   <p><img src="photo1.jpeg"></p>
    <!-- automated page, we don't know what this image actually is,
         it could be a TV screen capture, or a web cam on someone's
         head, or a photo someone just uploaded, we just don't know
         at this point and have no alternative text to give -->
   <p><img src="separator.png" alt=""></p>

...is better than this:

   <p><img src="separator.png" alt=""></p>
   <p><img src="photo1.jpeg" alt=""></p>
    <!-- don't have good alt text, so we'll leave it blank -->
   <p><img src="separator.png" alt=""></p>

...because in the first case, the user agent can easily ignore the two 
images before and after, and can just tell the user about the middle 
image, whereas in the last case, the user agent has no way of 
distinguishing the first and last images from the second (actually 
important) image.
   


> 4) <video> - On one hand you point to stats that show that LONGDESC is 
> misused and/or so poorly used that it should be dropped, and then you 
> suggest that most authors will write:
> 
> 	<p><video src="monkey.mpeg" controls></video></p>
>       <p><a href="monkey.mpeg">Download the Monkey video</a>.</p>

I didn't suggest they'd write that. I merely said that we want to allow 
this, and that this is why we don't want to make it non-conforming for the 
<video> element to have no fallback content.


> Respectfully,

With all due respect, invoking deities ("what on God's green earth"), 
telling me that you "need proof", telling me to "*PROVE*" that your 
approach is worse than mine (especially without doing the opposite), 
demanding that I answer your e-mails, asking rhetorical questions ("I 
mean, you did consult somebody, right?"), etc, is hardly respectful. It 
would help greatly if you could moderate your tone from its current 
aggressive stance to a more pleasant, cooperative one. We are working 
towards the same goal: making the Web a better place for everyone, 
including people with disabilities. Assuming that I am incompetent at that 
task and treating me as such is not a productive way of working together.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

Received on Friday, 7 September 2007 00:06:21 UTC