Access Matters

Since publishing the previous post, a fellow has stepped up and volunteered to continue Access-Matters. The domain name and the existing content of this blog become Anthony Ettinger’s later today.

Anthony Ettinger is a web developer and UI engineer with an interest in accessibility.

Anthony may or may not continue hosting the blog’s content. He may or may not continue adding new content. Those are his choices.

Thanks again to all who I have met here. Best wishes in all of your work and your other interests in life.

It’s been over two years since I retired from my last accessibility employment. I kept this blog around just in case I decided to continue accessibility consulting, go freelancing, etc.

My stint in accessibility consulting brought me many wonderful friends, colleagues and acquaintances. It also gave me the opportunity to help improve the web a little bit for people with disabilities, especially inside one very large corporation. Since retirement, I’ve watched accessibility matters from the sidelines, but spent more time pursuing other interests. I won’t rule out something causing me to reengage in accessibility work, but it’s not happening at the moment.

The domain name is up for renewal in early March. I’m going to let it lapse this time.

THANKS to all I have known in the accessibility world. It has been a very rewarding pleasure to know and work with you.

Coming out of retirement briefly, I write to support Glenda Watson Hyatt who yesterday lamented that WordPress still doesn’t get the very first accessibility requirement right. Bluntly, at release 2.8.4, WordPress still does not do ALT text correctly. Bloggers inserting images into their WordPress blogs are not being well served. Good craftsmen everywhere are taught to blame themselves for poor work rather than their tools. In this case, we can blame the tool. It needs sharpening.

Maybe I can cast Glenda a helping hand, not that she needs any help, by approaching the problem from a different point of view, from that of standards compliance, but tempered with explanations that are more easily understood than the W3C specs. In a comment to Glenda’s post, I suggested reading the specs. Yet, people using the WordPress tools are not always developers and shouldn’t have to know or understand those specs.

My intention here is to describe those specs in plain English, and to make suggestions for how WordPress can improve.

The discussion involves two “elements,” IMG and CAPTION, and two “attributes,” ALT and TITLE. Think of “elements” as containers, or envelopes. They are uniform carriers for what is inside them. Think of “attributes” as things inside the containers which offer additional precision.

IMG element

The IMG element is, of course, the container for images. Inside we find a pointer to the image and a number of allowable attributes. For the sake of completeness, the attributes are: src, id, name, alt, lang, title, style, onclick (and 9 other on… attributes), ismap, usmap, longdesc, align, width, height, border, hspace, and vspace. Only two of these are pertinent to our WordPress discussion, alt and title.

alt attribute

The alt attribute is intended for providing an alternate description of an image. The intent satisfies two cases. For browsers that do not or can not display images, the alt text describes the thing that is not seen. The other case is to give blind people a description of what they cannot see. The alt attribute is currently a REQUIRED attribute. It MUST be specified. Beyond the basic HTML specifications, the World Wide Web Consortium’s (W3C) Web Content Accessibility Guide also makes alt attributes for images REQUIRED.

Yes, there are exceptions, but before we get to those, here is what happens to the alt attribute. In browsers that are unable to display images, or have images display turned off, the alt text is used inside brackets or parenthesis to describe the missing image, for example (picture of dog rescuing swimmer). People who are blind use assistive software called screen readers. As the name implies, this software determines what is on the screen and “reads” it through voice synthesis to the blind person.

You can easily find long discussions about what images deserve alt text, which ones don’t, and what constitutes good alt text. Briefly, the only images that do not need alt text are those that are used for positioning, the ever present “spacer” gifs, and those that are purely decorative. Even the picture of the dog rescuing the swimmer might be considered decorative by some. Yet, even with these non-essential images the alt attribute must be coded and should be coded as alt=”". That is, two immediately adjacent quotes with no intervening spaces.

All other images need the alt attribute to describe the image. For the images such as drawings, illustrations, photos and charts, the most succinct brief description is appropriate. For images of buttons or other objects representing actions, the description of the action is appropriate, for example “submit button.” There are many ways to do alt text either very well or very poorly, but I don’t intend to go into those here. See references at the bottom of the article for good suggestions.

Oh, some readers may be aware that the next version of HTML, version 5, might not require the alt attribute. That is really not yet settled and the current draft of HTML 5 offers a variety of conditions. Those interested can read the draft.

title attribute

The title attribute is intended to provide additional information about the image. Title attributes are optional and can be applied to a great many elements other than image and are more generic in their nature. In our simple case of the dog rescuing the swimmer, the title text might be “Spot saving Dan.”

Many of today’s browsers display the title text when one hovers over an element that has a title attribute. A point of historical reference is useful here. The hover capability came from an implementation mistake that Microsoft made with Internet Explorer many releases ago. In that browser, Microsoft displayed the alt attribute as hover text. That was corrected later, making title the source of the hover text. It’s little wonder that people are confused by these.

The point is that title should be used for additional or advisory information, not as a substitute for alt text. As an example of advisory, think of a form full of fields. Some of those fields might have a title of “required information” to encourage correct completion.

Titles are not substitutes for the alt attributes and should not be used as such. Nor should we use the same value for both alt and title. That introduces redundancy that is very annoying for a blind person listening to a screen reader. In fact, because of so many developers who chose to irrationally duplicate alt and title text, many screen readers now have title announcement turned off by default. The redundancy became so annoying that blind people said they would rather do without the title information, even when it might be useful.

CAPTION element

Caption? Why are we talking about caption? The reason I ask is because the CAPTION element is an independent element that is intended to be used with TABLES. In fact, the specification is very precise about this, saying “When present, the CAPTION element’s text should describe the nature of the table. The CAPTION element is only permitted immediately after the TABLE start tag. A TABLE element may only contain one CAPTION element.” From this we know with certainty that the CAPTION element is optional, is associated with tables, and most be coded in a certain sequence for tables.

It is not an attribute for the IMG element and nowhere in the W3C specifications are captions ever associated with images.

On the other hand, some people do like to add caption-like information to pictures. In some cases, it might be a photo credit. In other, it might be yet more information about the image. However, that wasn’t what the HTML designers were thinking when they wrote the spec. Is it fortunate of not that most browsers are forgiving enough to not care about a caption outside of the correct place inside a table specification? The forgiving nature of browsers led developers to use the element for purposes other than those the designers intended.

Where WordPress gets it wrong

Following is the dialog for adding images to WordPress blog entries.

Let us count the problems:

1. There is no way to specify ALT text. Actually, there is, but it is well hidden.
2. The Title field is marked as required when in fact the attribute is optional.
3. To make it easy for people to do the wrong thing most of the time, the Title field is pre-filled with the image’s file name. More often than not, file names are not useful as either title or alt text.
4. There is a Caption field which is valid only for tables, not for images.
5. UNSEEN – If no caption is specified, the Title value is duplicated into the ALT text. I’m guessing the developers though most people would recognize what “title” means, but not what “alt” means, and blindly duplicating is easier than explaining.
6. UNSEEN – If the caption is specified, the caption’s value is duplicated into the ALT text. This is even farther form the spec than using the title text.
7. There’s no help for what any of the fields mean.

For the person in a hurry, who cares little about the details, we end up with what Glenda calls “Worse than Useless.” We have identical alt and title text, both specifying the image’s, usually cryptic, file name.

Ah, but there’s more….

In a very obscure effort to do things right, there’s an Advanced Settings dialog that includes a real, honest to goodness, field for alt text. If you really, really, really know what you are doing and really, really, really want to do it right, you can add an image, and then go to the Advanced Settings dialog to make the alt text correct.

However, getting to it is difficult. First, you have to complete the “Add an Image” dialog by pressing the “Insert into Post” button. Then, you have to find the image in the post and click on it. The, click on the button to “Edit Image.” After that, find and select the “Advanced Settings” tab.

Let’s recap. WordPress has an underlying understanding that alt text is required for images. However, they adopted the very poor practice of thinking title text more important and auto filling alt text from the title text. Then, a couple of releases ago, someone came along and injected the caption element into the mix, both misusing it and abusing it to auto fill alt text. To those of us acquainted with the standards, these combinations, especially the use of optional attributes to auto fill required attributes, are bizarre.

Lastly, there’s yet one more accessibility related standard that has been ignored. The Authoring Tool Accessibility Guidelines talk specifically about how the tools used to build web pages need to be accessible themselves, but also need to generate standards compliant code. WordPress fails here by not doing the right things with the “Add an Image” dialog, and by hiding the right things somewhere people are unlikely to discover.

Suggestions

The fix is simple. Do it right. Include a field for alt text. Make it the required field. Let title be an optional field. For complete standards compliance, get rid of the caption field. Yet, I can understand why some would want to keep it. Maybe the Advanced Settings dialog could be eliminated to make things easier? Lastly, and perhaps as important as all the other changes, provide good help information that informs the typical blogger about why alt text is required and what constitutes good alt text.

For further reading:

• WebAim has an excellent guide: Appropriate Use of Alternative Text
• As does Simon Willison at the Guild of Accessible Web Designers
• and Trenton Moss, writing for WebCrecible
• Current HTML 4.01 Specification
• Web Content Accessibility Guidelines
• Authoring Tool Accessibility Guidelines
• Draft for HTML 5
• Jez Lemon’s excellent article, Requiring the alt attribute in HTML5

A visitor, Ben Boyle, recently wrote that he was surprised when OSX Voiceover announced a series of three periods as “elipsis.” It is a surprisingly accurate interpretation of a simple character string.

That got Ben to wondering about how other characters, and special encodings, are announced.  Ben put together a selection of special characters and the variety of encodings that are commonly used for those characters. He provided the core of this case.

We agreed to try two different character set encodings, the older but ever present ISO-8859 and the broader UTF-8. I can already see significant visual rendering differences between the two character sets, with UTF-8 being the more accurate. It will be very interesting to know how the screen readers handle the test.

We are interested in knowing how theses things are announced by as many screen readers as you folks can use.  Please listen to either or both of the following test cases. The tests are rather lengthy, but I think well worth while. Take notes on what you hear, and respond to this posting with your results.

• Special Characters on a page coded as UTF-8
• Special Characters on a page coded as ISO-8859

Preface: Visitors who have been here before might remember that I once worked as an Accessibility Consultant for IBM. I retired early this year after an extraordinary 40 year career with IBM. What follows are my own opinions and should not in any way be attributed to IBM.

There was no “settled law” when the suit was brought against Target, no legal precedent to stand upon. That was likely one of the strong motivations for the National Federation of the Blind to bring the suit; to force settled law, and to clarify (actually greatly extend) the Americans with Disabilities Act. There’s still no settled law and I think that’s a very good outcome!

Those who think the settlement did not go far enough raise three concerns:
1) No legal precedent is established. This concern is also stated as the ADA’s applicability to the web wasn’t confirmed.
2) The settlement for class action litigants should have been larger.
3) The accommodations are mostly for the blind and not other disabilities. Jared Smith says, “Additionally, there is no indication that the Target accessibility guidelines include anything that would benefit anyone except blind users.”

My views are very much different. I’ll address the concerns in reverse order, doing the easiest one first.

3) The accommodations are mostly for the blind and not other disabilities.

Well … the lawsuit was brought by advocates of the blind. What else should we expect? Lawsuits are very serious matters, intended to decide very specific, and often precisely detailed, disputes. The case was very specifically about access for the blind, not access for the deaf or those with motor skill or other disabilities. The settlement must legally answer the charge, and has no obligation to satisfy unspecified complaints.

Sure, we who know how to do accessibility right would suggest going further and satisfying other disabilities as well. If we were designing from scratch and had good reason to believe the full range of people with disabilities would visit the site, we should very strongly recommend accommodations that satisfy all. That’s not the case in a lawsuit. Specific complaint: specific settlement.

Having dealt with solving these problems, I’m know that satisfying the blind is one of the toughest disabilities to accommodate. I also know that many things done to help the blind also help other disabilities as well. While Jared Smith sees solutions for only the blind, I see help for many others as well. For example, making everything keyboard operable helps the blind, those with motor skill problems, and all the rest of us who choose to be keyboard centric. Applying orderly document structure and paying attention to good tab order helps not only the blind, but also people with certain types of dyslexia, as well as all the rest of us who appreciate plain simple order. Ensuring that forms completion has a logical tab order also benefits the dyslexic and all the rest of us. If any specific disability has been left out, it might be the deaf. Yet, I don’t know how much multimedia (needing captions) actually exists on the Target site, or is planned.

2) The settlement for class action litigants should have been larger.

Let’s look at the details.

First, there are numerous costs for reworking the site, and a payment schedule for having the site monitored for compliance. Yes, I’m sure Jim Thatcher will provide very good value for the money, but compliance verification should have been left to Target to decide on their own, not a part of the settlement. After your last reckless driving and speeding judgment, you might have been sentenced to safe driving school, but has anyone heard of having a policeman assigned to monitor your driving behavior for the next three years. Wouldn’t it have been enough to say, “We don’t want to see you back in court again” and leave it to Target to decide how to monitor their compliance?

Then, there’s $20,000 to the California Center for the Blind on behalf of the original litigant. I don’t know what the suit actually cost the Center or the litigant, so I won’t quibble with that.

The remainder of the $6M is for class action claimants, awarding them $3500 if the site failed for them once and $7000 if they had more than one failure.  $3500 is a lot of money for not being able to buy something. I have no idea how legal “damages” are calculated under the California Unruh law. Trebling is a fairly common practice. Yet, I have a hard time believing anyone would have a $1200 cash register receipt from Target. I have no idea what the average receipt is at Target, but it seems beyond belief that the typical customer averages much more than $350 per visit. Using $350 would make the compensation ten times the amount that someone might have purchased. It also looks as though the compensation was designed to satisfy about 1200 possible claimants (about the mid point between 850 claimants at $7000 and 1700 claimants at $3500).

My view is that $3500 is very generous compensation for the inconvenience of having to go somewhere else to shop. Consider the case where your local Target store was late opening one morning, or had a broken water pipe that kept it closed all day. You drove over and couldn’t buy that patio furniture you had been wanting. Should you get $3500 compensation for not being able to access the store? Sure, it’s an artificial scenario, but it’s simply not normal for people to be compensated for being unable to purchase something, especially when there’s a K-Mart down the street that carries essentially the same merchandise (and just might have an accessible web site).

Then, Jared Smith says, “I also believe the 6 million dollar settlement to be rather insignificant for a corporation that had $63 billion in revenue and $3 billion in net income in 2007.” Oh please! Are there still people who believe in punishing the big guys just because they’re big? Robin Hood politics doesn’t belong in this settlement. Just as the complaint was for a specific class of people, the blind, the compensation should be specific, relevant to the complaint, not to what a corporation earns.

1) No legal precedent is established. This concern is also stated as the ADA’s applicability to the web wasn’t confirmed.

Jim Tatcher says, “The importance is now evident. In my non-legal opinion, the Judge said that the ADA applied to Target.com (to the extent that the web site related to Target’s bricks and mortar stores) and, most importantly, that the California anti-discrimination law (referred to as the Unruh Act) applied to Target.”

Peter Abrahams says,  “The laws on accessibility apply to your site.” and, “The law is now clear and public…”

Not at all!

Yes, those statements would be true had the case gone to trial and Target had actually been found guilty. Target was allowed to settle without admitting any wrong doing. The case never went to conclusion. Judge Patel’s ruling that the ADA and Unruh apply was not legally determined. It is still unsettled case law, still as murky as it has always been. Just like previous attempts, none of the cases came to a conclusion sufficient to claim the ADA is applicable. Three notable attempts have all failed to uphold the ADA as applicable to the web. (Gumson vs. Southwest Airlines, Spitzer vs. Ramada and Priceline, and NFB vs. Target)

The ADA became law in 1990, well before the advent of the web. The accommodation parts of the ADA are about physical accommodations, and there is very specific language and guidance about what types of physical accommodations must be made; parking spaces, aisle widths, braille markings, wheel chair ramps, rest room components, and on and on. The ADA material is so specific about these physical accommodations that one has to have a very fertile imagination to think that any of it could possibly apply to the web, or any other electronic systems. Applying the ADA to the web is nothing more than wishful thinking, thinking that requires numerous tenuous extensions to get from physical buildings to electronic enterprise. This looks like yet another case of a court inviting a penumbra into its judgment.

I understand why people would like to see the ADA applicable. It is the closest existing U.S. legislation that has the power to punish non-governmental businesses for how the disabled are treated. Yet, it is obsolete when the Web is considered. The right thing to do is to modernize the ADA through legislative action, not through judicial action. Create new law that is appropriate for non-governmental electronic enterprise. Make it as specific for the Internet as the ADA has been for physical accommodations. Pass it and enforce it. Stop trying to turn old law into new with judicial activism.

My conclusion

I still believe as strongly in making accessibility improvements as I always have. Businesses will be very well advised to avoid the legal hassles and lost opportunities that Target, Ramada, Priceline, Southwest Airlines, AOL, and others have experienced in the past.  Yet, I’m not sure many will see these as strong enough lessons to call to action.

The real messages to businesses is that full accessibility is simply good for business. While I worked for IBM, one of my primary responsibilities was looking after the accessibility features of ibm.com. The corporation was dedicated to providing full accessibility, and simultaneously was also very understanding of the competitive advantage that comes from being more accessible than the competition.

Free enterprise thrives when burdensome regulation and governmental interference is at a minimum. Let the businesses decide how they want to operate. Assuming equally attractive lines of products and services, those businesses who ignore some customers will suffer, and those who invite all will flourish. The last thing any of them need is a bludgeon based on a legal whim instead of clear and accurate law.

Hidden text can be a spammer’s haven.

Years ago, when screen readers were at release 4.0, they started to change and actually pay attention to display:none. I ran a series of tests back then and knew exactly how every screen reader behaved. I was one of the earliest advocates of hiding helpful text off to the left of the visible viewport. I’ve used it cautiously for things like skip links, for the real text associated with image replacement headers, and for occassional other situations where the blind could benefit from having text available that wasn’t part of the visual design.

Along the way, people speculated whether search engine owners would eventually scoff at the technique. It is too similar to the nefarious methods of “link stuffing” that many spammers practice. I ignored the issue … until yesterday when I received the following from Google:

Dear site owner or webmaster of access-matters.com,

While we were indexing your webpages, we detected that some of your pages were using techniques that were outside our quality guidelines, which can be found here: http://www.google.com/webmasters/guidelines.html

In order to preserve the quality of our search engine, we have temporarily removed some webpages from our search results. Currently pages from access-matters.com are scheduled to be removed for at least 30 days.

Specifically, we detected the following practices on your webpages:

* The following hidden text on access-matters.com:

e.g.

cation approval card credit instant Low apr interest credit cards Juniper credit card application. Lowest phentermine prices Credit score Apply for a fleet low interest credit card application Zithromax no prescription Zoloft interactions Credit credit card applications? Hoodia information Propecia effectiveness, Reporting and interpreting liabilities Credit scores explained! Zocor lipitor Side effects from fosamax medicati

[...]

Ah gee, has it gotten so bad that they have to crack down on my very simple use of “off left?” Didn’t they look to see that the class is “access,” making it clear that it’s done for accessibility? So, it was off to their guidelines documents, and to their forums. The guidelines were not specific enough to answer my questions. The forums had lots of discussion, but most was speculation by curious people like me … and no answers from anyone with Google credentials. The best information I found was an interview with Google’s Matt Cutts at the StoneTemple Consulting site. In that interview, Matt said:

So, our philosophy has tried to be not to find any false positives, but to try to detect stuff that would qualify as keyword stuffing, or gibberish, or stitching pages, or scraping, especially put together with hidden text.

We use a combination of algorithmic and manual things to find hidden text.

There was little to lead me to believe they had cracked down hard. Yet, I was skeptical that they might no longer be doing the “manual things” and making due consideration. So, lets’ have a look…

I pulled up the site and disabled CSS using Chris Pederick’s excellent Web-Developer toolbar. ARRRRRRRRRRRRRRGHHHH! Just below the header area was a huge block of spam links. What the ??? Where the H___ did they come from?

Long story short, the site had been hacked. The hacker left behind one line of code in the header.php file. That line started with <font style=’position: absolute;overflow: hidden;height: 0;width: 0′> and contained about 150 links for various pharmaceutical products, doubtful financial products, and enhancement products of all types for all genders. Yes, the hosting service confirmed the password thefts. Yes, I cleaned it up and changed my passwords.

Yes, Google was right. Spam was stuffed into hidden text on my site.

They were not complaining about my accessibility technique.