Re: summary="" in HTML5 ISSUE-32

On Feb 26, 2009, at 8:16 PM, Ian Hickson wrote:

> On Thu, 26 Feb 2009, Matt Morgan-May wrote:
>>>
>>> We're not just removing it, we're replacing it with something that
>>> helps more users (<caption>).
>>
>> ...which fewer authors will use, since it forces a visual
>> representation.
>
> I think that the total number of pages that use <caption> in a  
> useful way
> after <caption> has been advocated for for a decade would be higher  
> than
> the total number of pages that use summary="" in a useful way today.
> However, that _is_ just an opinion and I don't know how to prove it  
> either
> way.

I don't think that is really relevant to the discussion. This s not a  
contest between 'caption' and 'summary'.

> In general though, I think it's clear that features that are visible  
> to
> most authors have been used more appropriately than those that are not
> (e.g. href="" is used more correctly in general than cite="").

But there is no reason we should assume the 'summary' attribute is not  
visible, simply because it has not been in the past. If we need a  
visible feature we should work to make it visible. There's no point in  
creating a whole new feature and making that visible when we can just  
make the feature that already works visible where and in ways that  
make sense (like though user CSS).

>
> [snip]
>
>>> I don't agree with either of these statements. My position is just
>>> that there is data showing that summary="" as designed both fails to
>>> help disabled users (by being mostly bad data when used)
>>
>> I looked at the data posted most recently, and I came to a different
>> conclusion. For one thing, summary="(null|whitespace)" is what we've
>> used to indicate layout tables for years. And as has been mentioned
>> today, the biggest real issue is the mistaken practice of stating
>> "layout" or "design", and assistive technology across the board has
>> heuristics to filter that out. You're basing your decision on a  
>> problem
>> that in practice does not exist.
>
> Given Philip's data set, for instance:
>
>   http://canvex.lazyilluminati.com/misc/summary-20090226.html
>
> Could you point to a summary value that isn't bad?

Haphazardly clicking on 7 sites, all 7 used the summary attribute as  
advised. The first  6[2] used summary="" to indicate a layout table.

This site[1] included:
summary="The Details about World Bank U.S. Affairs displayed below"
Which I did not need as a visual user because a quick glance at the  
table oriented me to what was following.
  • http://www.wantagefirestation.co.uk/flashes/intro.html

So this is not a scientific study, but it does nothing to convince me  
that there is widespread misuse of the 'summary' attribute. Could you  
point to some sites where the attribute is not only misused but also  
harmful to an AT user.

> There are a couple, maybe, but the vast majority fall into one or  
> more of
> the following categories:
>
> 1. Summaries for layout tables; those tables shouldn't be there in the
>    first place.

This might be your own personal view, but it does not constitute the  
current state of HTML recommendations. You seem to include it here  
only to add support to your claims. CSS has replaced the need for most  
layout tables, but I think that AT and other UAs handle occasional  
layout tables just fine and they do not really do the harm that the  
old-school layout tables did with spacer gifs and all that mess.

> 2. Summaries that are actually better as captions.

I found none that would necessarily suggest that. If an author makes a  
conscious decision to leave out a caption because it is not  
necessarily needed for visual users but provides the information in  
the summary attribute I think we should be commending such an author  
and not criticizing their work. If the summary and caption are  
redundant than that is another story, but one that can be easily  
handled by UAs.

>
> 3. Keyword stuffing for abusing search engine ranking algorithms.

That's a slight problem, but an AT user will quickly recognize that  
without much harm done. Can you point out some examples though?

> 4. Completely unhelpful text.

Again can you point out some examples? How was it unhelpful?

>
>
>
>>> and hurts non-disabled users (by causing there to be information
>>> hidden from them when the attribute _is_ used in a way that helps
>>> disabled users).
>>
>> ...except that to date, the content of the attribute is explicitly  
>> for
>> the benefit of non-visual users. Any analysis of the existing data in
>> coming to this conclusion is invalid, since visual users were never
>> intended to receive it.
>
> That misses the point. The summary="" attribute might be intended  
> only for
> non-sighted users, but that doesn't mean that it's ok for the  
> information
> in it to be hidden from sighted users even when the information  
> would be
> useful to them.

Typically it shouldn't be useful to sighted users. However, there's no  
reason we as spec framers should be assuming that the 'summary'  
attribute must not be available to other users. Why do you keep making  
that inappropriate assumption.

> Consider the exact opposite case: <canvas> and <img> are intended  
> only for
> sighted users. Does that mean that it's ok for the content in those
> elements to be hidden from non-sighted users? No! We have to convey  
> the
> information from those elements to _all_ users, hence <canvas>  
> fallback
> and alt="".

This is a tortured analogy. A better analogy would be to say that the  
'canvas' and 'img' element are not available to users of aural user  
agents, so shouldn't we not allow those features at all. Like Maciej's  
earlier claim that providing features like the 'summary' attribute  
violates our design principles, this is totally turning the design  
principles on their head: claiming that violating the design  
principles is not adhering to them and that adhering to them violates  
them. Media independence means that we will have some features  
supported by some media and not others, but that the document provides  
a complete semantic encoding of the authored document: one that can be  
presented on various media with the most minimal loss of information.

> If summary="" is causing people to hide useful information from  
> sighted
> users -- as Steven's data showed -- then we should find a way to  
> encourage
> authors to expose the information to _all_ users. The HTML5 spec today
> does this using <caption>.

I don't see how you thought Steven's data showed that. Can you say  
what you mean by that? The 'caption' element is fine for captions. It  
is inappropriate for summaries. We should be sure to convey that to  
authors. But if authors are using 'summary' for captions we haven't  
been shown evidence of that. If authors were using 'caption' for  
summaries I'm sure that would be a common complaint from web users.

> [snip]
>
>>> New pages that use <caption> wouldn't be any less accessible either.
>>
>> Where's the proof of that?
>
> The example for <caption> in the spec is one existence proof.

Actually this statement proves the opposite. If authors of complex  
tables believe they have made them accessible simply by providing a  
caption, they will tend to make tables less accessible.


[1]: <http://web.worldbank.org/WBSITE/EXTERNAL/COUNTRIES/EXTUS/0,,menuPK:668561~pagePK:64168427~piPK:64168435~theSitePK:668549,00.html 
 >

[2]

* http://aackbar.online.fr/virginieledoyen/sommaire.php3
  * http://www.shopin.co.il/
  * http://www.kingston.org.au/
  * http://www.soroschool.org/
  * http://www2.warwick.ac.uk/services/musiccentre/