W3C home > Mailing lists > Public > public-html@w3.org > August 2010

Re: providing a long description using the summary and details elements.

From: Leif Halvard Silli <xn--mlform-iua@xn--mlform-iua.no>
Date: Tue, 24 Aug 2010 18:50:18 +0200
To: Steven Faulkner <faulkner.steve@gmail.com>
Cc: HTMLWG WG <public-html@w3.org>, HTML Accessibility Task Force <public-html-a11y@w3.org>
Message-ID: <20100824185018213823.8a302131@xn--mlform-iua.no>
Steven Faulkner, Tue, 24 Aug 2010 16:04:09 +0100:
> Hi Leif,
> 
>>I would add <details role="img"> as a requirement, whenever <details>
>>is used to represent an image. Justification below.
> 
> it's not being used to represent an image its being used as a 
> container for a long text alternative.

We have been here before ... The ARIA 1.0 definition of role="img" is 
"container": A container for a collection of elements that form an 
image. There is nothing here which says that one of the elements 
inside the container cannot be a button.

That said, if you have your concrete example in mind, then perhaps you 
would like to suggest a better role than "img"? The image contains a 
diagram, what's the appropriate role for a diagram?

>>Making @longdesc a boolean like that, does not make sense to me, for
>>several reasons. One thing is that it would go against how it is
>>currently specced (and what about those bogus @longdescs out there?).
>>But just as imporant: to rely on @longdesc like that, would be to not
>>take advantage of <details role="img">,
> 
> i have suggested that
> " longdesc attribute with URL value (longdesc="URL") provided for 
> backwards compatibility (suggest deprecated in HTML5)"

A double meaning then? HTML5-compatible UAs see it as a boolean 
attribute, even when it contains a URL, while HTML5-incompatible UAs 
see it as a link? 
 
> also
> "> - use of longdesc attribute on image provides indication to browser/AT
>> that details/summary is being used for the specific purpose of
>> providing a long description."
> 
> how many bogus longdesc are there to be found in <summary> elements?

I know. But I don't think it is smart to introduce one interpretation 
inside <summary/> and another elsewhere. Plus that old UAs probably 
would see an empty, boolean longdesc as equivalent to longdesc="", 
which they - again - would interpret as a link to the current page.

However, before we resolve whether such a reuse of @longdesc could 
acceptable, it is important to know why a @longdesc of the boolean 
kind, should be is needed.? In <details/> there is already a relation 
between <summary/> and content. And your example keeps the <img> in the 
<summary/>, and thus I don't understand what a boolean @longdesc then 
can add - compatible user agents already know that the content of 
<details/> is some kind of long description of the summary inside 
<summary/>.

Perhaps, what you suggest, is that we shall have two categories of 
<details/> elements: those which are used for "normal" interaction. And 
those where the content contains a long description of the <summary/>? 
What kind of benefits would such a division in two categories provide?

>>One thing is that it would go against how it is currently specced
> 
> how is that an argument?

Before I can comment more on that, I need to know the answer to why 
such a re-interpreation of @longdesc should be needed. THe problem I 
have is that new, HTML5 compatible clients will understand <details> 
and <summmary>. Thuse these user agents doesn't need an additional 
boolean longdesc. 
 
>>But just as imporant: to rely on @longdesc like that, would be to not
>>take advantage of <details role="img">,
> 
> I was trying to provide a native HTML5 method without recourse to 
> ARIA, besides i don't agree that it should be role="img".

Again: What @role do you suggest, in your concrete example, then?

>>If the <details/> element has role="img", then everything outside the
>><summary/> element naturally becomes the long description of what is
>>summarized in the <summary/> element. Hence, there would be no reason
>>to introduce @longdesc as some kind of boolean signal - such a thing
>>would only be redundant. (And sometimes even incorrect - imaging if
>>both @alt and text outside the @alt makes up the <summary/> text.)
> 
> As i said I don't agree that the details element should have a role="img".

In your concrete example? Or in general?
 
> your code example:
> <details role="img">
> <summary><img longdesc="link" alt="Short desc"></summary>
> <a href="link">long desc link</a>
> </details>
> 
> problem with this is that everything inside the details element is 
> guven a role="img"
> so the summary element, which will probablyu be exposed as a button 
> to AT will become an image, so will the link.

So, are you saying that <details/>, unlike a <figure/> cannot 
technically be presented as an element with role="img"?

My code example is otherwise "thought up" in analogy to the generic 
example in ARIA 1.0: [1]

]]
<div role="img" aria-labelledby="caption">
  <img src="example.png" role="presentation" alt="">
  <p id="caption">
   A visible text caption labeling the image.</p>
</div>
[[

The above ARIA example is, from AT user's angle, equivalent to this:

<div role="img" aria-labelledby="caption">
  <p id="caption">
   A visible text caption labeling the image. 
   <img src="example.png" role="presentation" alt="">
</p>
</div>

Which is equivalent (for AT users) to this:

<figure role="img" aria-labelledby="caption">
  <figcaption id="caption">
    A visible text caption labeling the image. 
   <img src="example.png" role="presentation" alt="">
</figcaption>
</figure>

When it comes to <details>, then we give increased attention to the 
caption element for _all_ users (not only to AT users). Thus the above 
<figure/> example can be turned into a <details> element with a "long 
descripton", like this:

<details role="img" aria-labelledby="caption">
  <summary id="caption">
    A visible text caption labeling the image. 
   <img src="example.png" role="presentation" alt=""
    longdesc="detailsbody">
  </summary>
<p id="detailsbody">
 Long description text of image ...
</p>
</details>

(PS: I hope the above example can help illustrate why it is unhelpful 
to not have the same caption element for both <details/> and <figure/>, 
as it is often useful to switch from <figure/> to <details/> and 
vice-versa.)

>>Suggestion: the ARIA equivalent of role="iframe". 
> there is none, and again I would like to see a native solution, 
> rather than one that requires the addition of ARIA.

I would not describe ARIA roles as an addition but merely as an 
description. An ARIA role is an ARIA representation of something in the 
host language. ARIA 1.0 has (at least) this to say about frames:

]]
Superclass Role: section
Subclass Roles: alert, article, grid, landmark, list, log, status, 
tabpanel
Related Concepts: HTML Frame [ snip ]
[[
 
>>One of the @longdesc use cases in
> HTML4, is providing information about an iframe element.
> 
> this may have been useful in the earley days when AT did not support 
> <iframes>, to be honest I have never had cause to recommend using 
> longdesc on an iframe.

OK. 

Coming back to your proposal to use <iframe> as the conten of a 
<details> element: I doubt that it will become generally acceptable, if 
the the purpose is to provide a long description of of something 
summarized in <summary/>, if the iframe  content is provided primarliy 
for AT users, as I belive many developers will consider it quite 
"expensive" to load an external page  - to all users - for that 
purpose. A (longdesc) link would then be considered cheaper.

[1] http://www.w3.org/TR/wai-aria/complete#presentation
-- 
leif halvard silli
Received on Tuesday, 24 August 2010 16:50:57 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 24 August 2010 16:51:00 GMT