W3C home > Mailing lists > Public > public-html@w3.org > February 2008

Re: ISSUE-30 longdesc Re: Clarification of rational for deprecation...

From: Charles McCathieNevile <chaals@opera.com>
Date: Wed, 06 Feb 2008 17:41:53 +0530
To: "James Graham" <jg307@cam.ac.uk>
Cc: "HTML WG" <public-html@w3.org>
Message-ID: <op.t53qt3z2wxe0ny@pc130.chandigarh.osa>

On Wed, 06 Feb 2008 15:39:39 +0530, James Graham <jg307@cam.ac.uk> wrote:

>> Could someone with access to the issue tracker please raise an issue  
>> for whether longdesc should be deprecated?

> Done. Since your name was one of the options for "Raised by" I selected  
> that.


>> Actually, I disagree on this. In our own research into about a million  
>> pages we came up with a low usage of longdesc - a bit less than one per  
>> cent of pages. And of that usage, something like half to two-thirds was  
>> totally useless.

> (It would be interesting to have a little more information of the  
> methodology used in your research, e.g. how you chose the sites)

Ergh. It's a collection we had internally for this purpose. I don't know  
how it is chosen, but the results are likely to be skewed to western  
sites. However, I think our actual numbers (I don't have them in front of  
me now) are broadly in line with Ian's and other similar studies. In any  
case, my point is that the actual numbers aren't that critical. In my  
opinion the interpretation that those numbers justify deprecating longdesc  
is in flawed.

>> ... Of the few users who rely on it, or benefit substantially from it  
>> ... most are in a position where something that works 20% of the time  
>> is better than something that doesn't work at all, since they do not  
>> get an alternative.

> I think this 20% number is misleading in describing how often longdesc  
> "works"; unless I have understood incorrectly, 20% represents, in terms  
> of a conditional probability

No, the 20% figure (being measured as P(is useful | exists) was given as a  
number that is even lower than the figures people have suggest. In other  
words, even if the situation with markup is much worse than we think, then  
the overall effect of the longdesc attribute is actually more beneficial  
than has been claimed (and the impact of abuse is far less negative, as  

>> There are a number of cases where it would be easy to provide longdesc.  
>> One example is sites that provide walking or driving directions both as  
>> a map and as a textual explanation of the route (yes, I do ask blind  
>> friends to get me driving directions to somewhere and blind people do  
>> give driving directions to their house, favourite pub, etc). While I  
>> suspect most site authors will not do this in the coming 5 years,  
>> helpful usage of the attribute has certainly increased dramatically  
>> over the last 10 despite there being no way to access its content for  
>> half that time.

> I don't understand why longdesc is useful here. If the site provides  
> both text and a map (which it should because the text is beneficial to  
> everyone) there is no obvious requirement for an invisible indicator  
> that the text is an alternative to the map; the screenreader can jut  
> ignore the map and read the text.

Except that the screen reader doesn't know which bit of the text replaces  
the map. Try reading your screen through a drinking straw for an hour to  
get an idea of the limited context that a blind user has. You simply  
cannot scan the entire page at a glance with a screenreader - you have to  
listen to the structure of it carefully and remember all that, or read  
through the entire thing to find stuff, unless there are explicit  
associations such as longdesc.

This efficiency problem is the reason why I really care about semantics -  
in general, I don't think that people are very good at using any kid of  
semantics and simple things like header markup have required massive  
amounts of effort to get half-right.


> However, in this case it's not at all clear to me that @longdesc does  
> benefit users because it's not clear that encouraging the alternative as  
> best practice (solutions involving visible content accessible to all  
> users) wouldn't help the people who supposedly need @longdesc more often.

In many many cases the content should indeed be visible. However, the  
longdesc should flag the association of a piece of visible description and  
the thing that it describes (which is somthing you can easily and readily  
do with @longdesc today, or ten years ago). And in some cases, the content  
will not be visible because it upsets the design flow, but you still need  
a way to point to it - again, something that longdesc achieves.

Longdesc is not a perfect solution. But in terms of design it seems a lot  
better thought out than alt - it may be badly used at least as often, but  
were well used it is able to improve more in more cases...



Charles McCathieNevile  Opera Software, Standards Group
     je parle français -- hablo español -- jeg lærer norsk
http://my.opera.com/chaals   Try Opera 9.5: http://snapshot.opera.com
Received on Wednesday, 6 February 2008 12:12:23 UTC

This archive was generated by hypermail 2.4.0 : Saturday, 9 October 2021 18:44:26 UTC