Rebuilding The Web

Articles, advocacy, discussion and debate about the many problems of the Web and the challenges of rebuilding it.

How do we save longdesc?

The longdesc attribute, although potentially useful, was removed from the HTML5 specification, despite recommendations to retain it from the HTML Accessibility Task Force. The decision to drop longdesc was a lazy response on the part of the HTML5 team: the attribute isn't popular, so let's ditch it. At the same time, the Task Force's attempts to simply reinstate the longdesc attribute won't suddenly make the attribute popular, or ensure it is used as intended. What both parties need to do is remember their obligations. Once they do that, a proper solution will emerge.

The HTML5 team's obligations

The HTML5 team focuses on how HTML is used right now. On that basis, while they recognize the value of longdesc, they perceive it is unpopular and practically never used as intended. But they neglect to ask why this is the case, and how the problems with longdesc can be fixed. They see their responsibility as limited to how HTML is used, instead of how it should be used.

Accessibility expert obligations

The responsibility of accessibility experts/advocates goes beyond reinstating the longdesc attribute into the HTML5 spec. They also need to change user/vendor attitudes toward this attribute, to make sure it fulfills its considerable function. To accomplish this, accessibility experts must look at accessibility from the point of view of the content author and ask why the longdesc attribute is so unpopular right now.

A possible solution

As a lead developer of a WYSIWYG editor and as someone who often interacts with content authors, I believe the longdesc attribute is not used as intended because the longdesc attribute and the related alt attribute are poorly defined in the HTML specification. Consequently, derivative works such as articles, references and tutorials propagate the incorrect use of these features. Influenced by such works, browser and tool vendors then implement these features inconsistently and haphazardly.

Many content authors (incorrectly) understand that the alt attribute contains short descriptions of images, while the longdesc attribute contains longer descriptions. From the author's perspective, this is of course confusing. The author asks: why do I have to write something short, then have to duplicate my efforts by writing something longer? The short/long description explanation simply does not work for content authors.

Instead I suggest we start describing content that appears in the alt attribute as "textual substitute" for an image, and content appearing in the longdesc as a "description" of the image. It's a subtle but meaningful difference that makes a real distinction between these two attributes, instead of it being simply a difference in the number of characters typed. The consequences of such a change would be significant for accessibility and greatly assist authoring tool vendors in creating better interfaces with meaningful labels that will encourage authors to write appropriate content. It's not too late to fulfill the promise of the longdesc attribute. Update articles, HTML references, and tutorials on the distinction between alt and longdesc and how each should be used. Show browser vendors how alt and longdesc should be rendered, and demonstrate to tool vendors the benefits of creating better user interfaces.

Conclusion

Simply reinstating the longdesc attribute into the HTML5 spec will not make the Web more accessible. Instead, it is the duty of the HTML5 team and accessibility experts/advocates to make accessibility features such as longdesc popular with users. One way to accomplish this is to define the alt and longdesc attributes in a way that is meaningful to content authors and easy to implement for browser/tool vendors.

Public comments

1. Posted by Laura
on Sunday 2010-08-22 at 17:18:09 PST

Bugs had been filed to improve longdesc. For instance:

longdesc URL checking:
http://www.w3.org/Bugs/Public/show_bug.cgi?id=10015

Native user agent support for exposing longdesc to all users
http://www.w3.org/Bugs/Public/show_bug.cgi?id=10019

All have now been marked WONTFIX or INVALID by the editor.

Related References:

ISSUE-30: include a longdesc attribute for images - Straw Poll for Objections
http://www.w3.org/2002/09/wbs/40318/issue-30-objection-poll/results

HTML5 Working Group Decision on ISSUE-30 longdesc
By Sam Ruby, Maciej Stachowiak, and Paul Cotton
http://lists.w3.org/Archives/Public/public-html/2010Aug/att-0112/issue-30-decision.html

Notice of Impending Formal Objection to HTML5 Issue 30 Decision (@longdesc)
http://lists.w3.org/Archives/Public/public-html-a11y/2010Aug/0027.html

Longdesc is Dead! Long Live Longdesc!
http://www.cfit.ie/news-and-commentary-archive/525-longdesc-rip

Strategic Decisions in a Strategy-less Environment
http://burningbird.net/node/118

Podcast #83: Fate of Longdesc in HTML5
http://webaxe.blogspot.com/2010/08/podcast-83-fate-of-longdesc-in-html5.html

Alone in the Pitch Black Dark
http://www.cssquirrel.com/2010/08/16/comic-update-alone-in-the-pitch-black-dark/

No longdesc Attribute in HTML5
http://www.456bereastreet.com/archive/201008/no_longdesc_attribute_in_html5/

2. Posted by mattur
on Monday 2010-08-23 at 05:54:46 PST

> "The responsibility of accessibility experts/advocates goes beyond reinstating the longdesc attribute..."

Richard Schwerdtfeger, Distinguished Engineer in Accessibility at IBM, and Chair of the W3C's WAI-PF's WAI-ARIA team, has previously described longdesc as "a disaster" [1] that should be "buried in the back yard" [2]

I get the impression he may possibly think longdesc is somewhat suboptimal, and may even disagree with you about his responsibility to reinstate it.

[1] http://lists.w3.org/Archives/Public/public-html/2009Sep/0596.html
[2] http://www.w3.org/2009/11/03-pf-minutes.html

3. Posted by Vlad Alexander
on Monday 2010-08-23 at 06:57:04 PST

mattur wrote: "Richard Schwerdtfeger, Distinguished Engineer in Accessibility at IBM, and Chair of the W3C's WAI-PF's WAI-ARIA team, has previously described longdesc as 'a disaster'..."

I think that everyone agrees that longdesc has not worked in the past. However, past performance is not a good criteria to remove a feature from HTML. The criteria should be 'can this feature be made to work in the future?' I believe it can. As a WYSIWYG developer, I can imagine a user-friendly interface that can encourage not only the use but also correct use of longdesc. I cannot say the same for aria-describedby for images (which Richard Schwerdtfeger mentioned).

FYI, if HTML5 team was consistent in their decision making process and removed features based on low popularity and incorrect use, many more features would have been removed such as q. But authoring practices can change. Until a few years ago, blockquote was used primarily for formatting, which is not the case today for newly authored content.

4. Posted by The Pachyderminator
on Monday 2010-08-23 at 20:00:36 PST

I can't see that the solution you propose would be beneficial, because it's hard for me to think of a situation where a verbose description of an image, apart from the alternative textual content, would be necessary or useful. The cases where longdesc would be useful are ones where usable alternative content should be something more than plain, unmarkedup text such as the alt attribute allows. For example, the most useful alternate for a graph would be a table, not a long, repetitive paragraph.

For this reason, I would consider the longdesc attribute (if properly supported) to be an alternative to alt. Images should be required to have only one of the two. For images that can be adequately replaced by fairly brief plain text, longdesc isn't useful, and for images that need something more - whether a longer text that ought to have semantic markup, or a table or something of that kind - alt isn't.

5. Posted by Vlad Alexander
on Tuesday 2010-08-24 at 11:49:46 PST

The Pachyderminator wrote: "I would consider the longdesc attribute (if properly supported) to be an alternative to alt. Images should be required to have only one of the two."

We have been incorrectly trained to think that alt and longdesc are essentially the same thing. They actually serve quite different purposes. longdesc makes images accessible. alt makes content surrounding the images comprehensible when images cannot be seen. Since you cited the example of a graph, in the following, what should be the alt and longdesc?

Sample line graph.

The longdesc can contain the following data, which is useful to both blind and sighted users as well as machines such as search engines:

Annual Sales
YearBillJoan
1990$500$200
1991$550$250
1992$550$300
1993$500$400
1994$550$500
1995$500$600
1996$550$700
1997$600$750
1998$600$800
1999$650$850

However, you cannot determine what the appropriate alt content should be without taking into consideration its context (i.e. how the alternate text works with the text that precedes it and the text that follows it).

If the text preceding the graph reads: "By default, my graphing program creates graphs that are unattractive and difficult to read", appropriate alternate text would be something like: "The background is grey, gridlines are black and the series lines are pink and navy."

Suppose instead the text preceding the graph reads: "To convert the following to a bar graph, click on it to select it, then press the Chart Wizard button on the toolbar; in the 'Chart type' field select Bar and press the Finish button." Appropriate alternate text in this case would be something like: "Line graph."

If the text preceding the graph reads: "As a reward, Joan got a big bonus when her sales surpassed Bill's for the first time." Appropriate alternate text would be: "This happened in 1995."

Suppose the text preceding the image reads: "Between 1990 and 1999, Joan's annual sales increased from $200 to $850 while Bill's were stagnant, ranging from $500 to $650." In this case, since the preceding text adequately describes the purpose of the image, alternate text could be left blank (empty alt attribute), or it could be a general description such as: "Graph of annual sales for Bill and Joan in the 1990's."

So we can see from these examples that we need to start teaching that content appearing in the alt attribute serves as a "textual substitute" for an image, with its function being to make content comprehensible when the image cannot be seen. This is quite distinct from the role of content used in longdesc, which simply describes the image in its own right.

6. Posted by The Pachyderminator
on Thursday 2010-08-26 at 08:57:27 PST

Vlad wrote: "We have been incorrectly trained to think that alt and longdesc are essentially the same thing. They actually serve quite different purposes. longdesc makes images accessible. alt makes content surrounding the images comprehensible when images cannot be seen."

Point taken, and thanks for elaborating the example of the graph, which makes your argument, as far as it goes, seem incontrovertible. However, it does not take into account the large number of cases where the image itself and the content conveyed by the image are one and the same. This will be easier to discusss with a real example: http://starwarsmodern.blogspot.com/2010/08/architecture-of-inception-combat.html. This is a long article that has an image after every paragraph or two. The content of the article is perfectly comprehensible and without gaps even with all the images removed. Does this mean they ought all to have valueless alt attributes?

However, it would be inaccurate to describe the images as mere eye candy for sighted users. Each one is a juxtaposition of two (often seemingly unrelated) pictures, and often the purpose is to point out the similarities or differences between them. In each case, you could write a sentence expressing the main idea, but if you simply used those sentences as alts, the result would be unsatisfactory, because each one would stick out like a sore thumb and obviously be randomly scattered, and in some cases would have little relation to the content immediately before or after. The fact is, writing those sentences at all would be pointless. The only way to accomplish the purpose of the images in text would be to describe the images at length. In this and and hundreds of analogous cases, the distinction you make between the funtion of alt and that of longdesc breaks down.

7. Posted by Andrew Downie
on Thursday 2010-08-26 at 16:02:19 PST

I am a screen reader user. Sadly, in my many years of using HTML-based resources, the only examples of correct use of longdesc I have found are those I have created for demonstration purposes. I occasionally come across longdescs and they invariably link to the current page. That isn't to say that they couldn't be useful, but a great deal of work will be required if that is to happen. My understanding was that a longdesc is only available to screen reader users and not to visual readers. Is that incorrect?


Andrew

8. Posted by Vlad Alexander
on Thursday 2010-08-26 at 21:23:56 PST

The Pachyderminator, I don't think you can have both "the content of the article is perfectly comprehensible and without gaps even with all the images removed" and "each one [image] is a juxtaposition of two (often seemingly unrelated) pictures, and often the purpose is to point out the similarities or differences between them".

If the images do not add to the comprehension of the article, then the alt attribute should be blank.

If the images do add to the comprehension of the article, then the alt attribute should not be blank unless surrounding content adequately describes the purpose of those images and alternate text then becomes duplicate content.

The Pachyderminator wrote: "The only way to accomplish the purpose of the images in text would be to describe the images at length."

I don't believe so. The author of the article expects a certain level of knowledge on the subject matter from the intended audience. If an image, say of a movie scene, has meaning to the reader, then textually identifying that movie/scene is adequate alternate text. One does not need to write the main content of the article for one audience (those with knowledge of the subject matter) and then have to dumb-down the alternate text to suit another audience group.

Andrew Downie wrote: "My understanding was that a longdesc is only available to screen reader users and not to visual readers. Is that incorrect?"

The screen shot below shows how Opera supports longdesc:

Context menu for an image contains the option "Long Description".

9. Posted by Dave
on Friday 2010-08-27 at 11:52:15 PST

I agree that dropping longdesc removes a feature that's needed, if underused and misused. In every discussion I've seen, focus has centered on HTML solving the problem.

Could something like "longdesc" be stored in the image file?

Think about that. There's already some metadata in most image file formats, I think. Since the longdesc data is so closely tied to the image file, it makes more sense to store them together in one file rather than two separate files that both have to be managed, linked to, verified, updated separately, etc.

There would be some times when one image needs different longdesc depending on its use, but I think that would be the exception. The ease of use of storing them together would outweigh that negative.

It could be somewhat of a bandwidth waste for someone who needs longdesc but does not need/want to download the visual image. Perhaps if the metadata is stored at the beginning of the image file, it would be possible in this case to just request the first few packets of an image file, and stop the download once all the metadata is retrieved? (If screenreader setups typically download the whole image file, this might be a moot point at the moment.)

10. Posted by Vlad Alexander
on Saturday 2010-08-28 at 08:17:57 PST

Dave wrote "Could something like 'longdesc' be stored in the image file?"

I think this is feasible for future image file formats.

11. Posted by Andrés Sanhueza
on Tuesday 2010-08-31 at 11:03:26 PST

Do you know if there are cases of a "paragraph + image" content where would be preferable to have a single alternative replacement for both, instead of just one for the image (and if there's a way to achieve that)?

12. Posted by Vlad Alexander
on Wednesday 2010-09-01 at 06:42:38 PST

Andrés Sanhueza wrote: "Do you know if there are cases of a 'paragraph + image' content where would be preferable to have a single alternative replacement for both, instead of just one for the image"

Nothing comes to mind.

13. Posted by Denny Lancaster
on Wednesday 2010-09-29 at 04:12:49 PST

If the images do not add to the comprehension of the article, then the alt attribute should be blank. Which can also be said of the long file description. But if either or both apply, then a developer should actively question why the image was included in the first instance.

While a greater part of discussion is placed concerning blind users, the simple fact that other disABILITIES such as comprehension, critical thinking and short and or long term memory retention (or the lack there of) are significantly enhanced or reinforced with graphical representations.

So the lack of use of either the alt attribute and or the long file description boils down to: a lack of understanding or correctly stated "empathy" and just plain laziness. Probably a combination of both.

Comments are closed for this article.

Main menu

Check out the a11y bugs project that aims to help browser / tool vendors fix accessibility bugs.