Transcription: Web Axe Episode 83 (Fate of Longdesc in HTML5)

[Introduction, woman's voice over music] Welcome to Web Axe, practical web accessibility tips. Web Axe dot blogspot dot com. Web Axe. Web site accessibility. Web standards. Web Axe dot blogspot dot com.

Dennis: Hello and welcome to Web Axe podcast number 83. This is your host Dennis. And today we have a special episode. We are going to be speaking with a few web accessibility professionals about the longdesc attribute and if it will remain in HTML5. There has been some talk recently on the web and Twitter about the fate of longdesc and if it's needed or not, and if it should be in HTML5 or not. So this is a pretty good discussion. So listen in. This is myself and John, Everett and Joe. Enjoy.

OK. So this is Dennis from Web Axe and we have Joe Dolson, Everett Zufelt, and John Foliot on the line to speak about the controversial issue of the longdesc image attribute and its fate in HTML5. So first off, let us have you guys introduce yourselves briefly. Joe.

Joe Dolson: I am Joe Dolson. I am accessibility consultant and web developer and I write a monthly column for 'Practical eCommerce Magazine' on accessibility issues and ecommerce.

Dennis: And Joe, you've been on the show, but it has been quite a while, so welcome back. Thank you. Everett.

Everett Zufelt: Yeah, I am Everett Zufelt in Toronto, Canada. I am an accessibility consultant and developer and I am also the Drupal accessibility maintainer.

Dennis: Excellent. Welcome. And you are also helping with some hosting for accessibletwitter. So thanks.

Everett: You are welcome.

Dennis: And John.

John Foliot: Good morning Dennis. John Foliot, I too am an accessibility consultant and specialist. I am currently working at Stanford University as well I am actively involved in the W3C Accessibility Task Force for HTML5 where amongst other things I am the co-chair of the sub group working on accessibility of media. Dennis, I'd like to if I can just to make it very clear that anything I say today is my own personal opinion and does not reflect the opinion of others within the W3C. Just to get that out there.

Dennis: Right, understood. Thanks for clarifying. And yeah, if you recognize John's voice he was recently on the show and speaking we discussed HTML5 and accessibility. So welcome back John.

John: Thanks Dennis.

Dennis: OK. So just to give a real basic background so everyone knows, or hopefully that the alt attribute is what you add to an image tag to make it accessible. So you enter a brief description on that image, if it is not decorative, so that users who were unable to see that image for one reason or another can still have the meaning conveyed to them.

So there is this kind of a limit and that area is fussy too and how long and there was a recent blogging discussion about it. The link was circling around Twitter that our post in the show notes. So the length of it is relatively short. If you have like a complicated image, say like an image of a diagram or something requires more than a sentence or two description, then you need a longer description.

So that is why in the HTML4.01 specification there is something called the longdesc attribute. Where you enter a URL which points to a long description of that image so you can fully describe what's in that image. So now the issue is HTML5 and there is a question on whether that longdesc attribute should be included or not. So John, can you tell us the state of that right now and what is going on?


Dennis: Well just the basic stuff, like is it in the spec right now and...

John: Earlier this week on Wednesday the chairpersons of the HTML5 working group reviewed all of the evidence that was put forward and they made an evaluation that the longest attribute would be removed from the HTML5 specification. So at this point in time, the decision is that it will not be in the spec.

There has been, as you have already noted a fair amount of controversy about that and there has been a number of indications that decision are going to appealed. As a matter of fact a gentleman by the name of Leif Halvard Silli who is a Norwegian member of the group has already filed a formal appeal which will be, works its way all to Mr. Tim Berners-Lee himself appealing the decision.

This appeal was based on a number of issues that are for the most part procedural but nonetheless legitimate and there are other appeals that are being worked on behind the scenes right now. So there will be multiple appeals on various reasons.

Dennis: OK. So well let's start the discussion then on that. So Joe, can you say however briefly you can put it like, if you agree or disagree or you are kind of in the middle, as to whether it should be included or not?

Joe: Well I do think it is somewhat of a difficult question because I'd never felt that the longdesc attribute was really the best way to convey complex information that's in an image. Ultimately I feel that these highly detailed images are themselves a problem. The information should be conveyed in a way that's more globally available.

And there have been a lot of advances in that with these alternate methods where people have created scripts that will convert your table of data into a graphic, into graph of some sort and that's I think more helpful on the whole. But whether the attribute should actually be in the spec I find to be complicated. And I think there has been a certain amount of distraction from the frankly specious argument that it should be removed because it is not used. That is an argument which is really quite empty.

Dennis: Yeah, hardly anybody uses the caption element on a data table all right, but that doesn't mean it shouldn't be used or that we should take it out, because it obviously should be there. But anyways, OK. Everett.

Everett: Well, early in the week I was quite indifferent and so I decided to have a few conversations with different people about the topic. And I guess where I fall at this point in time is that there needs to be some mechanism. Earlier you mentioned that the alt attribute was to give a brief short description of an image. And the current HTML5 tag actually explicitly states that the alt attribute is not to be used to describe the image. The alt attribute is to be used to give a textual equivalency when the image isn't available.

And so the equivalency is not necessarily going to be a description. It is going to be the meaning the author or the person who has put the image into the document, the meaning they wanted the image to convey. And so I think that we have two different textual representations of an image that we need to have a mechanism to deal with. The first is the textual equivalency or the semantic meaning of the image and the second is a more detailed description of the image. There needs to be mechanisms to have both of those pieces of content available when the image itself is not available to the user.

Dennis: OK. So, the question that keeps going through my mind is... I agree that the longdesc attribute isn't really helpful. But that's only because in my viewpoint, that for one, developers or designers haven't implemented it correctly. Two, the browsers haven't implemented very well either the functionality of it, or supported it at all. So, if those two things were better, would you have a more favored opinion of the longdesc attribute? Do you agree that that's a problem, Joe?

Joe: The longdesc attribute, in my mind, is simply a method of providing access to this information, and I don't think that the method is really what's important. What's important is providing access to that information and that detail.

John: I'll only half agree with that Joe.

Joe: What was that?

John: I'll only have agree with that, but continue.


Dennis: Don't worry John; I'll get back to you in a second.

Joe: My opinion is that, it's very important for accessibility to have a requirement that images with complex information are described. And, I agree that I don't think the attribute is a good place for that. My question is whether the longdesc attribute is necessarily any better then just using a figure caption, or anchor, to point to more detailed information, or just having that text on the page. I think it needs to be required, I'm not clear that the longdesc itself is the way to do that.

Dennis: OK. Everett.

Everett: Yeah, I like Joe's approach, which is, let's start with the problem that we're trying to solve first. So, longdesc traditionally through implementation, and also, probably just through the way that it can only be implemented by pointing to an external source for the content, likely is not the best way to solve the problem that we have.

What we need, I would say, is a method through which the author can point whether it is to an external resource, or a resource on the same page. And make some sort of connection saying, over here on the page is a description, here is an image and I want someway to be able to programmatically link those two things together, like I'd link a label to a form field, or in the way that a figure had a figure caption, and there's a very explicit relationship between the item, or the object, and the textual record [inaudible 12:51] .

Dennis: Yeah, and I agree with that. I think if maybe HTML5 could extend the longdesc attribute so that you could point to the page, point to the content on the page, as well as pointing to external content, that might help.

John: You can today.

Dennis: Yeah, well... [Laughter]

John: You can today.

Dennis: Let me just point out real fast before I turn it over to you John, that I don't know if you're referring to that aria-labeledby attribute, but in ARIA, which is supposed to be a bridging technology, as the W3C states themselves, you can do that. But still, that doesn't solve the problem for the actual HTML5 specification where, I think, we all agree, there needs to be some kind of improvement on this issue. So, John.

John: OK. So, a couple of things that have been said that I'd like to address. First of all, I think the method that we do it is important. Right now we've got two ways that we can associate a longer tract of text to an image, so that we can actually describe what the image is. One is longdesc, and the other is aria-describedby.

Aria-describedby, as Dennis points out, is part of the suite of ARIA tools that is intended by the W3C's own admission to be a bridging technology. aria-describedby can link an asset, in this instance an image, to a tract of text that is on the same page. So, it can point to what's called an ID ref, or what we also think of as a named anchor, so we have intra-page navigation. What you can do with longdesc, by the way. longdesc equals pound placed elsewhere on the same page, and that will do exactly the same thing.

So, we've got two mechanisms. The problem with aria-describedby is, a: it was intended to be a bridging technology, but b: and more importantly, it can not point off page. It can only point to something on the same page so, in that regard, it's perhaps, a little bit weaker. I think, the more important thing and the thing that needs to be discussed is the quality of what we're pointing to.

One of the reasons why those that are in favor of obsoleting the longdesc, is that they're saying that the quality of the text that's being pointed to is of little to no value 99.999% of the time. In other words, it's garbage. That's an argument that's hard to counter, because it's true.

A lot of people, especially in the very early days, really didn't understand what longdesc was, really didn't understand how to do it, and very often, didn't do it even though they should have. So, we have a body of existing work, that if you go on the long tail going way back to 1997, when longdesc was first introduced, there's a lot of garbage out there.

But you know what? If I start looking at alt text from 1997 and 1998, there's a lot of garbage there, too. I challenge anybody to go and find pages from that era, and you'll for example, you'll have left hand navigation where the alt text is button, button, button, button.

So, part of the problem is also educating authors about the importance of the value that we're pointing to. It really confuses me that we a: are removing a native attribute that's been around for more then a decade now, in favor of an alternate technique, the aria-describedby technique, which is not as powerful as the longdesc attribute.

I mean, look, the basic building block of what we're doing on the web today is that, we link things to other things. I mean, that's what the web is, it's a link, linking things to other things. Longdesc is a very eloquent solution that links an alternative piece of text to a sophisticated image. The mechanism itself is not broken, it was how people have been using it in the past that we have a problem.

Dennis: Yeah, and that's kind of the point I brought up in the beginning is, how people were using it, not unlike other HTML attributes. Also, it was disappointing to me that the browsers kind of fell off in their support of it.

So, Joe, I mean looking backwards, I think the need to not... It's difficult to take away some of these items because some people do use it. It's included John informed me previously that longdesc is included in some laws, some countries laws. And a lot of these like Dreamweaver and other HTML editors support it and stuff, so what do you think Joe, about if the attribute was removed don't you think that would cause any problems or do you think looking ahead is more important?

Joe: I think absolutely it causes problems. My issue with it being removed isn't so much that I feel that longdesc is a critical attribute. I don't feel they're providing any kind of alternative. I mean, legally it would be messy but if there was an alternative that they were proposing to replace it which was equal then you could map your laws and say "Instead of this, we're going to use this now."

I don't want to get into the actual legal consequences of doing that, because that's quite messy, depending on the situation...

John: Years!

Dennis: [laughs]

Joe: ...but at least it would be possible. But I don't feel that they're actually proposing anything as an alternative. John mentioned that replacing it by aria-describedby by, but actually... I mean, when I'm looking at the document on the HTML5, where they list the items they're making obsolete. All they suggest is using a regular anchor to link to the description. And I don't think that is at all sufficient, because it doesn't any programmatic identification with the image itself.

Dennis: Agreed.

John: At least aria-describedby, though more limited than longdesc, does convey that association.

Dennis: That is true. Everett?

Everett: Yeah, I think I'd have to agree with everything that Joe said right there. And maybe to add more. I think we mentioned the fact that the W3C itself talks about ARIA as being a transitional technology. The reasoning behind that, though I don't know it exactly... My reasoning behind it is: it's not appropriate for HTML5 as a spec to rely on the web acceptability initiative, and ARIA, in particular in this case, to provide accessibility or DBAI.

If there's a need to make something accessible, for instance, longdesc or aria-describedby, that needs to actually be baked into the HMTL5 set. You shouldn't point off and say, "Here's a completely separate specification that you can use to make your HTML5 accessible." We need accessibility to be universal within the HTML5standard.

John: Everett, if I can though... I mean, part of the way that the W3C has been looking at HTML5, is more of a modular kind of an approach. So, a lot of things are being split off into separate or secondary documents, so that they can be worked on asynchronously from the larger document.

So, for example, HBG is in its own specification; RDFa is in its own specification. And so, having the ARIA specification as both stand alone but referenced from the HTML5 document is actually a good design choice, because that way if we need to make to the ARIA spec. We don't have to reopen the HTML5 specifications and vice-versa.

So it's the sum of the whole that is important. And so, I agree that we need to have an HTML5, as many tools as we can in the tool box, to enhance and improve on the accessibility of documents. But where that specifically is written down and stored, in what specification, I think is secondary.

Again, the nature of the web is re-linked to things. So having the HTML5 document linked to accessibly guidance from the Center of Expertise at the W3C is a good design pattern. It's not a bad design pattern.

Dennis: OK. Good point. I'm not sure where else to go with this. Joe, do you have any final comments?

Joe: Well, not really on the longdescing. I was interested by the slight deviation into design patterns of HTML5 and ARIA. I agree that it's good to have these separated patterns within the specifications. However, I'm inclined to think there should be a characteristic for this issue, the longdesc issue, within HTML5, simply because I don't think...

[inaudible 23:21] is the core of the specification. You should be able to create any document without necessitating the use of the other specifications for accessibility or other issues. You don't have to use SVG. But if you put in any kind of complicated image... Now, certainly you need to use aria-descirbedby or HTML recommendation to use a regular anchor element. But, I've already mentioned that that is really kind of inferior.

Everett: And add to that perhaps the semantic value of the attributes, in that, ARIA is meant to make accessible with Internet applications. And forgetting the difference between a regular document, in which I might have a complex image, or a rich Internet application, in which I might have an image.

If we forget about ARIA, there's still within the HTML5 spec itself no method of semantically linking a description to an image. And so, I think that maybe is a flaw.

Dennis: Yeah. And something that just pops in my head that's kind of obvious that we didn't mention is like the browser... Well, I mentioned the browser supporting lacking in longdesc, but that's a HTML specification.

But what about ARIA? I mean, if a browser is not supporting ARIA or that piece of ARIA, then we're really screwed, right?

John: Dennis, if I might... I think there's a light bit of confusion about this notion of browser support. So, in some testing that I did this past week - and admittedly only the PC platform for the testing. I mean, I've not yet cracked open a Mac to do it. But both the Opera and Firefox web browsers on the PC platform, if a sophisticated image has a longdesc, and you mouse over that image and then right click, you have a contextual menu that allows you to follow the links of a longer description.

So, we have native support for sighted users in both Firefox and Opera. However, the longdesc attribute and the value string that's associated to that attribute is stored in the DOM in all of the browsers. And so, for adaptive technology like screen readers that do support the longdesc attribute today. The two mainstream or the two most widely used screen readers being JAWS and Window-Eyes, both support the longdesc attribute.

So you might not have, for example, native support for longdesc in Internet Explorer 8. But for a non sighted user that's using JAWS or Window-Eyes in combination with the IE8 browser, they could still access the content of the longdesc attribute. So the support for the attribute in terms of the technology is actually a lot more widespread than a lot of people are prepared to admit.

Dennis: OK. I mean, I did a little bit of testing myself. I didn't see that in my Firefox but I did find that in Opera.

John: Depends on which build of Firefox you have.

Dennis: Yeah.

John: Mozilla has been waffling back and forth. I think I've got Firefox 3.6 or something like that. I'd have to check.

Dennis: OK. I mean, just generally speaking, I think overall support, and especially implementation, has been lacking. So if everybody could somewhat agree on something, then that would be good. Yeah, I think our conversation is... We may not have solved any problems, but I think we did a good job of exposing the issues and agreeing on some things where it's important.

We'll see what happens with the HTML5 spec. I myself, hope it stays in, although I think we all agree that it could use some help or maybe an additional method added in there. Yeah. But thanks for joining in the conversation, guys. And, hopefully, you could come on the show again sometime in the future.

Joe: Thank you.

John: Thank you.

Everett: Yeah, thanks.

[music and commercial]