Bruce Lawson’s personal site

Alternate text in HTML 5

For a while now there has been a battle raging between accessibility advocates and those who are most closely involved in drafting the HTML 5 specification.

I’ve bravely kept out of much of it as it’s been quite acrimonious and I am, as you might have realised, a delicate flower. (Actually, since I’ve been working from home on my own, I do take things much more to heart than I used to. Silly me).

The biggest arguments have been over the alt attribute for images. At some point, HTML 5 allowed alt to be optional; a huge flamewar erupted and now there is a massive lump of text in the spec discussing alternate text. I think this is wrong, and the spec should simply make alt mandatory (while noting that it may be blank) and point to WCAG for guidance; it is, after all, the authoring guide to accessibility from the W3C

On the debate on the summary attribute on table sees me side with my chums on the HTML 5 cabal. Now that we have ARIA attributes like aria-describedby, and HTML 5 elements like details (which “represents additional information or controls which the user can obtain on demand” – see the details element spec) I think there are better ways of doing it. (I feel the same about longdesc too.)

I’m suspicious of invisible screenreader-only elements. A long time ago, I worked with kids with learning difficulties—autism and Down Syndrome—and can’t help thinking that the information that’s usually in summary to provide a mental map of a data table to blind people would be as useful to those kids. (This discussion is taking place on Gez Lemon’s blog, in a very pleasant tone, I’m glad to report).

Anyway, yesterday Janina Sajka, the Chair of the W3C‘s Protocols and Formats Working Group pronounced on the two matters I discuss above in a document called WAI CG Consensus Resolutions on Text alternatives in HTML 5 which is worth a read (it’s neither long nor complex). I don’t know what happens now procedurally but hopefully it’ll put an end to those battles and everyone can have a lovely big group hug and a sing-song.

Until we get started on the accessibility of the canvas element, that is.

More of my brilliant observations on HTML 5 accessibility:

33 Responses to “ Alternate text in HTML 5 ”

Comment by Anne van Kesteren

Having looked at the document I am disappointed that it is just a list of recommendations we are supposed to take for granted(?) without providing any kind of rationale.

In my view it does not really address the critical questions that have been raised e.g. regarding longdesc but just states how that group of people think things should work.

On IRC this was characterized as A) “I’d like you to do X” B) “There are some issues with X, e.g. …” A) “I’d like you to do X”. That is no way of having a discussion in my opinion.

Comment by Bruce

Damn, there goes my sing-song dream, Anne. I had visions of Messrs Ruby and Hixie on guitar, with Ms burningbird, Foliot and Pilgrim singing.

Comment by John Foliot

One day, there is a knock at the door. Mr. Jones (not his real name) answers to see that it is telegram deliveryman.

“Oh,” says Mr. Jones, “You’re delivering a telegram, I want you to make it a singing telegram.”

“That might not be appropriate” replies the deliveryman.

“No,” says Jones, “I’d really like you to sing it for me”

“Well, answers the deliveryman, “If you insist.”
“Ta da, tada da da, You’re Mother’s Dead…”

The accessibility community has explained, offered use-cases, discussed, queried actual users, debunked myths and have thoroughly examined the alt attribute from every angle conceivable for well over 2 years now. It has listened to the contrary argument, and in the final recommendation has acknowledged possible shortcomings and situations when adding effective alt text is problematic, and offered guidance and recommendations for those use-cases as well.

Less than an hour after Janina Sajka, the Chair of the W3C’s Protocols and Formats Working Group released the official recommendation (a consensus document authored by no less than 4 W3C accessibility working groups), Anne van Kesteren dismissively rejects the hard work and sincere attempt to address a problem with a sarcastic toss off in the HTML5’s cabal IRC channel. For shame!

Yes, Anne, you are supposed to listen and learn from teachers, peers, and experts. You cannot know everything, whether you are a 20 something asshole or a 50 something asshole – the difference being of course knowing when to admit you might be wrong. For all your cleverness, this appears to be one lesson you’ve yet to learn.

The discussion has been had, the thoughts, issues, ideas and concerns of both sides of the debate have been weighed up, and after multiple groups concerned with web accessibility finished reviewing the data, their recommendation has been put forth. The real test of course will be whether the HTML5 WG accepts and implements the consensus document into HTML 5, or whether you and your buddies think you are smarter. A very large number of people are watching the outcome here.

Comment by Shelley

Bruce, me singing would be beyond cruel. Might be an effective weapon though. “Do this, or Shelley sings!”

Anne, I’m not sure what you mean by “rationale”. Reasons for the recommendation were given. Or do we need to reduce this back to some basic level: We need alternative text because blind people can’t see the image? At some point in time these demands become less an interest in clarification, and more a stubborn rejection of what you see to be “experts”.

It is OK for people to have expertise in a field. It is OK for people to know more about a subject than you. It is fine for people to be experts, and to give an expert opinion.

Can you provide a link(s) to the specific concerns raised about longdesc? Something that sums it up? This is for me, since I’ve seen some stuff, but not necessarily anything truly definitive.

And what problem do you have with the recommendation about alt?

Comment by Kyle Weems

On the negative side, it looks like the celebration of peace triggered more fights (Hi Anne! Thanks for ruining the group hug!)

On the plus side, if this keeps up I’ll have plenty of material for a new comic at some point. Of course, I’m thinking parodying it might just be throwing more fuel on the fire.

Still, can’t we just think of the people with genuine accessibility needs here?

Comment by John Foliot

@kyle For what it’s worth, I believe that the 4 working groups within the W3C *do* “…think of the people with genuine accessibility needs…” – heck many of those group members *ARE* members of that same group. But of course that is meaningless to the cabal brain trust who have all the answers (even when they don’t fully understand the problems)

Comment by Kyle Weems

@John – I guess I should have phrased that better. I know most of the associated parties are genuinely concerned. I applaud that majority. A portion, however, seem to object to experts other than themselves who are strongly opposing any ideas they didn’t invent.

I was directing the commentary towards that portion, and am dismayed at their choice to wear blinders.

Comment by Shelley

Never mind, Anne, I found the original WhatWG blog entry at

One response, and I made the same with summary, is that anything that came about because of Ian running tests against Google’s index is not a valid test, and amounts to nothing more than anecdotal information.

There is no third party access to either the original data, or the methodology used in the test.

Ian either needs to stop using Google data and resources as validation of his choices, or resign as HTML5 author because of conflict of interest.

Comment by Anne van Kesteren

Shelley, similar data has been found by “open” tools.

I am also not disputing the need for alternate text, I am disputing the specific recommendations in that document for solving the problem as they appear to come out of thin air for people who do not have access to the underlying reasoning, data, and debate.

I think an expert should be able to tell you why she/he is correct and does not have to state she/he is correct because she/he is an expert.

Comment by Bruce

There’s a third option, Shelley, that I prefer: open up the Google data that he has access to, so we can all see it.

A resource analysing a billion web pages seems like too valuable a resource to lose, and I don’t see any candidates other than Hixie to edit the spec (not that I know many spec editors).

@john: I’ve got nothing against calling people assholes (although on UK domain it should really be “arsehole”, but I’ll let you off cos you’re a nice bloke). But I think Anne’s point was genuinely made and calling names doesn’t exactly sit comfortably so close to the New Seekers.

Comment by Shelley

Bruce, I wouldn’t expect Google to open its index.

Anne, we’ve had discussions about past usage before. For instance with summary, we found that the table element wasn’t used correctly, which shows that much of HTML4 was probably not used correctly in the last ten years.

But what do you want in these documents? No better yet: what do you want from the accessibility folks that you don’t seem to be asking from yourselves, for other decisions made in HTML5?

For instance, if you want lab tests and scientific study, where is the lab tests and scientific studies for microdata?

What do you want, Anne? Give a specific example of what you feel is a “rationale”.

Comment by Matt May

When outsiders explain all the circumstances that surround users’ needs, and why therefore changes should be made to the spec, they’re accused of inundating the WG with “walls of text.”

When they break down the requirements as clearly as possible without defending every line item, they’re accused of opacity.

So, which is it?

The WHAT WG decided long ago that @longdesc failed because people can’t grasp using a URI to point to a long description. Post hoc ergo propter hoc. What makes them think that aria-describedby will suffer the same fate, when both it and aria-labelledby point to URIs, and when everyone using ARIA to date actually understands, implements and correctly communicates to others its role and datatype?

Comment by John Foliot

@bruce: my usage of that specific term was in direct reference to a comment Anne made on the IRC logs at: – so it is not so much name calling but rather referencing an identifier (and FWIW, I was referring to myself as the 50+ asshole – cause sometimes I am – and I know that I’ve been branded that term by the cabal in the past) However, if the shoe fits…

@matt: It’s just they think they have a much more elegant solution: A Caption. (Read description.) (

Ya, that’ll get a lot of buy-in: more code, rather than less, and “extra” text muddying up all those lovely Photoshop comps because “of those blind people”.

It’s almost as if every “accessibility” solution they propose is stacked in such a way as it will be completely ignored and omitted: honestly, it’s hard enough to convince people to add @longdesc or aria-describedby – do they really think we will be successful in having content creators add extra text and links to their creations – hardly. And what if there is more than one image that has Read description“? Why then we have multiple links using the same text identifier – creating yet another accessibility issue, not only for the blind, but also impacting on those with cognitive issues. Yep, they’ve clearly thought this out…

Comment by Mark

The “lack of rationale” that Anne is talking about (and that, as far as I can tell, no one else is talking about) is this:

“Can accessibility problems be solved with accessibility-specific markup that is intentionally hidden from everyone except the people it purports to help?”

This is the fundamental point of disagreement between the parties in this ongoing debate. It has been expressed many times, in many ways, in many forums. The Consensus Resolution does not address this issue; it simply assumes that the answer is “yes” and works from there. It is not addressed in the “Principles” section; it is not addressed in the “Context” section; it is not addressed in any of the “Specific Recommendations.”

There are other issues. The Consensus Resolution says “that HTML5 state that ‘For guidance on accessibility requirements for text alternatives authors should consult WCAG 2.0.'” As has been pointed out repeatedly, WCAG 2.0 offers woefully inadequate advice on how to actually author text alternatives in specific situations. HTML 5’s advice is much clearer, much more thorough, and backed by many more examples. Removing the text from HTML 5 and deferring to WCAG 2 would result in poorer text alternatives being authored by people who care enough to look for advice in the first place.

John Foliot can scream “THE EXPERTS HAVE SPOKEN” until the cows come home, but this does not advance the discussion in any way. We are aware of who has spoken, and we have read what they have to say. In some cases, these “experts” simply aren’t saying very much (WCAG 2). In other cases, they are not addressing the issues we have previously raised (Consensus Resolution).

Comment by steve faulkner

mark wrote:
“Can accessibility problems be solved with accessibility-specific markup that is intentionally hidden from everyone except the people it purports to help?”

Which part(s) of the WAI alt doc propose this? I suggest that the document proposes the opposite in regards to long descriptions in particular.

mark wrote:
“WCAG 2.0 offers woefully inadequate advice on how to actually author text alternatives in specific situations.”

This is a reason to improve and extend the WAI examples and techniques not a reason to write a competing set of text alternative guidelines.

mark wrote:
“Removing the text from HTML 5 and deferring to WCAG 2 would result in poorer text alternatives being authored by people who care enough to look for advice in the first place.”

Negative conjecture based on the idea that the WCAG 2.0 techniques for providing text alternatives will never be adequate. The WCAG 2.0 techniques are a work in progress, I am sure that WAI would consider the examples (where they do not contradict WCAG) provided in the html 5 spec as a basis for further WCAG 2.0 techniques.

Comment by steve faulkner

following up on another misleading statement I missed in my first response
mark wrote:
“Removing the text from HTML 5 and deferring to WCAG 2”

It does not say in the alt doc that text alternative examples be removed from html5 it says:

“HTML should not provide any guidance that conflicts with WCAG”

and that WCAG 2.0 be referenced “For guidance on accessibility requirements for text alternatives”

which is not unreasonable is it?

Comment by Mark

Steve, you’ve hit the nail on the head. The Consensus Resolution does not address the issue of whether hidden accessibility-specific markup is the right solution. It simply assumes that it is (as many other accessibility advocates have been assuming for over a decade now) and then goes from there. And that is precisely why some people are so disappointed with it. “Months of work!” And yet they didn’t even understand the objections that had been raised!

As for “not contradicting WCAG 2,” I assume that’s code for “make @alt required.” You can’t possibly mean it literally, since there’s virtually nothing in the WCAG 2 Techniques document about how to author useful @alt attributes. (If you haven’t read it, I suggest you do so. Trust me, it won’t take long!)

And if you want @alt to be required, then we’ll have to talk about what to do when no value is available. A nice concrete use case to discuss is, which auto-generates bogus @alt attributes for profile pictures because the original Twitter site does not provide any (nor does it provide any way for users to provide them). The Consensus Resolution seems to suggest the solution that has opted for (just making shit up). At least, that’s what they were doing the last time I checked, but that was a few months ago. Perhaps someone could check and see if they’ve solved this problem since then. God knows, we’d all like to find a solution.

Comment by John Foliot

Mark wrote:
“Can accessibility problems be solved with accessibility-specific markup that is intentionally hidden from everyone except the people it purports to help?”

I’ll ask Mark this question: when you encounter braille signage on an elevator, does it help anyone other than the blind who actually can read braille? Other signage might be printed, illuminated or applied in any number of methods, yet braille is always provided as a tactile presentation – simply because that is how it must be to work. Unless you read braille however it is effectively hidden from you (in plain site): does this mean then that it is wrong, bad, or inappropriate?

“The Consensus Resolution does not address this issue; it simply assumes that the answer is “yes” and works from there.”

The problem of course is that others, who have not spent any real time working with or in the communities directly affected here, *presumes* _no_ – they are, in effect, presuming to be subject matter experts with no experience or credibility to back that claim. Mark and his cronies might not agree with the starting point that the WAI Working Groups started with, but he fails to acknowledge that those working groups contain many people that *ARE* directly affected by the response – they *ARE* the community of users. Simply put, if a blind person says I need *foo*, and you yourself are not blind, what presumes that you can know better?

“WCAG 2.0 offers woefully inadequate advice on how to actually author text alternatives in specific situations. HTML 5’s advice is much clearer, much more thorough, and backed by many more examples.”

According to whom? Mark Pilgrim? It should be pointed out (and has been many times as well) that the examples provided for alternative text in the HTML5 draft are woefully incorrect: they are descriptions, not alternatives, and are often times overly verbose – don’t take my word for it Mark, ask the numerous blind users you have worked with in the past for their direct feedback. Those that I have asked are pretty much in agreement that they would not want to be hit hard with that much information every time they encountered an image: they want terse alternatives – longer descriptions belong in an external container. Those examples were written by one person, Ian Hickson, with zero input from any blind user (an admitted presumption on my part, but I have directly asked Ian on numerous occasions for a name, any name, of someone he collaborated with to arrive at those examples – since none is forthcoming one can only presume that it is because he cannot produce one).

“John Foliot can scream “THE EXPERTS HAVE SPOKEN” until the cows come home,…”

Mark Pilgrim can scream “WE KNOW BETTER” until the cows come home, but no amount of screaming that will make Mark a member of the community directly affected here.

Those directly affected by the question worked in a bona fide process to author a consensus document – a consensus reached by people with disabilities, people with technical backgrounds, and people with years or experience working directly in the field of adaptive technologies and delivery of content to disabled communities. Who do *you * think is better placed to comment: that broad community, or a bunch of techies that work at Google, Opera and Apple?

Comment by Leif Halvard Sili

This is a typical HTML 5 debate yes: 2 words of critique from Anne. And 200 words of defence from John. Kind of unbalanced.

In short: Anne is much too short (especially considered that his brings critique for lack of rationale …)

And I would recommend some of the “accessibilists” to try to dig out what e.g people like Anne means.

Comment by Leif Halvard Silli

Mark said: And yet they didn’t even understand the objections that had been raised!.

This sounds as if the WHATWG has some common and well worked out alternativeprinciples that it work from. I think you should write those principles down somewhere and also show how they could be applied. (In truth, the WHATwg often argue as if they are a bunch of agreeing persons, but as Mark and Anne demonstrate above, they have differing views [Anne said: I am also not disputing the need for alternate text, whereas Mark questions whether it is the right solution – while at the same time he claims to know what Anne meant.])

Mark talked about the issue of whether hidden accessibility-specific markup is the right solution.

And what is the alternative? That all accessibility-specific markup becomes visible?

We hear a lot of critisism of the Consensus document, but it is hard to find the coherent, well founded and principle based alternative proposal.

Comment by Michael Kozakewich

I think the comments are more exciting than the post!

For no reason at all, I installed the FireVox voice extension, and so it would belt out all kinds of words at me as I browsed.
Eventually, I decided it was annoying, verbose, long-winded, and almost unintelligable (though that last one was because it was Microsoft Sam), so I ended up disabling it.

I’d suggest that anyone working on accessibility options should spend a good chunk of time relying on those options. Try a screen reader out, and find out for yourselves how long is too long, and what kind of markup works best.
It really is hard to judge anything about it before trying it. If I’d been part of the team trying to put the spec together, I know I would have made some horrible errors.

Comment by Kyle Weems

@Eric & Bruce – I totally want to be in your non-cabal cabal. Are there ice cream socials? Because, let’s be honest, that’d be awesome.

I’m a non-expert in this field, but it sounds to me like John is asking a pretty sensible question. Has Ian, Anne or anyone else on their half of this discussion actually had discussions with blind users to confirm their needs and web usage desires/preferences? If not, they need to brake hard on the egos and accept that some outside advice on this might be important.

Comment by Eric Meyer

Great idea, Bruce! Only I can’t join because I’m already part of the shadowy microformats cabal. Nor can I resign in order to be part of your club– anyone can tell you that once you’re part of a cabal, there is no through, there is no out. It’s like Krazy Glue. So no non-cabal-cabal ice cream socials for me, I’m afraid.

I guess all I can do now is try to worm my way into the HTML5 cabal. Once I’ve done that, I’ll have to look around to find another one I can join in order to complete a set.

Leave a Reply

HTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong> . To display code, manually escape it.