Re[2]: CfC: Checkbox and Radio button labels and 1.3.1

Paul,

Just to be clear - the original thread brought up a couple of 
interesting issues.

1)  There are no examples of radio button or checkbox without a visible 
label that is programmatically connected to the input so that you can 
click on the label to check the checkbox or radio button.

2) You raised the issue that always having a clickable label is a 
desirable thing. In many cases the larger the better, as it makes these 
controls accessible to many users. All well and good, and no-one in the 
group would say this is a bad thing.

However, to comment on the second item first, and on clickable labels in 
general -  it just may not be desirable in *all* cases to mandate that 
developers do this - it may not be a suitable pattern depending on the 
environment, the way a form is put together etc. This is one of the main 
blockers to your original suggestion/issue IMO. So we are hesitant in 
suggesting that the group takes this direction. Yes, we have spoken 
about this at length on the calls and are discussing it here and the 
consensus supports our decision to leave things as they are. Again, no 
one is saying that per say, clickable labels are bad but they may just 
not be appropriate in all situations and our spec needs to provide 
gestalt guidance, and then as you drill down into our resources - 
specific suitable techniques that support a wide range of use cases.

As to the first point, I don't understand how a user is supposed to 
click on a non-visible label in the first place? Could you comment 
further on what you are thinking? Do you mean that even if a label isn't 
visible the 'hit area' may still be active as if there was a visible 
label?

Finally, on the first point, as Andrew points out we have existing 
examples of there inputs can be connected to visible label via 
aria-labelledby by and for/id combinations and even when there are no 
labels you can use the well supported @title attribute on <input> 
elements. All good. So no-one is arguing against programmatically 
determined relationships, and we are aware that we need to keep our 
techniques up to date as development and design patterns change (and are 
actively working on this).

So I look forward to more input from you on this, actually it would be 
great if you could attend a call to discuss - as this topic is 
interesting nuanced discussion - and may point the way to wider issues 
that the group need to deal with.

Thanks

Josh



------ Original Message ------
From: "Paul Adam" <paul.adam@deque.com>
To: "Andrew Kirkpatrick" <akirkpat@adobe.com>
Cc: "josh@interaccess.ie" <josh@interaccess.ie>; "Detlev Fischer" 
<detlev.fischer@testkreis.de>; "David MacDonald" 
<david100@sympatico.ca>; "Makoto UEKI" <ueki@infoaxia.com>; "WCAG" 
<w3c-wai-gl@w3.org>
Sent: 05/12/2015 00:45:59
Subject: Re: CfC: Checkbox and Radio button labels and 1.3.1

>All modern screen readers determine aria-labelledby properly, if not 
>let’s file a bug report.
>
>aria-labelledby is an explicit association between an element and the 
>id of another element whereas a checkbox and a text string inside the 
>same paragraph have no explicit association and I don’t see how they 
>could have a relationship just because they’re in the same paragraph. I 
>understand that passes for link purpose in context but I didn’t think 
>for info and relationships?
>
>Does that mean that form inputs with error messages below the input or 
>input format instructions don’t really need to be associated with the 
>error and info strings? They can just be in the same paragraph? Or in 
>close proximity?
>
>I did not think that you could claim WCAG conformance based on how good 
>of a guesser a particular screen reader is. I know that JAWS does lots 
>of guessing and VoiceOver does some as well whereas NVDA does not.
>
>I really hope we’re not promoting that these methods can pass WCAG!
>
>Thanks!
>
>Paul J. Adam
>Accessibility Evangelist
>www.deque.com
>
>>On Dec 4, 2015, at 4:22 PM, Andrew Kirkpatrick <akirkpat@adobe.com> 
>>wrote:
>>
>>Paul,
>>When using aria-labelledby which screen readers can determine the 
>>label of the checkbox?  Which ones determine this properly?  Of 
>>course, not all do (yet) and the way that you determine is to test it.
>>
>>Does the less-than-ideal code I suggested pass with all user agents?  
>>Undoubtedly not.  Does it pass with some?  Yes, and if those are the 
>>user agents that I use to base my accessibility support claim then 
>>that would be how I’d justify the pass.
>>
>>The relationship can be implicit as well as explicit and I believe 
>>that also includes the case where you have:
>>
>><input type=“checkbox” title=“Please send me a ton of email”> Please 
>>send me a ton of email
>>
>>I’ll re-emphasize that there is no doubt that using the explicit 
>>approaches are better, but the thinking expressed on the call I 
>>believe was that even though the other approaches are not as good that 
>>we can’t state that they fail.
>>
>>Thanks,
>>AWK
>>
>>Andrew Kirkpatrick
>>Group Product Manager, Accessibility
>>Adobe
>>
>>akirkpat@adobe.com
>>http://twitter.com/awkawk
>>http://blogs.adobe.com/accessibility
>>
>>From: "paul.adam@deque.com"
>>Date: Friday, December 4, 2015 at 16:55
>>To: Andrew Kirkpatrick
>>Cc: "josh@interaccess.ie", Detlev Fischer, David MacDonald, Makoto 
>>UEKI, WCAG
>>Subject: Re: CfC: Checkbox and Radio button labels and 1.3.1
>>
>>Hi Andrew, no this does not make sense to me.
>>
>><PastedGraphic-2.png>
>>
>><p><input type=“checkbox”> Please send me a ton of email</p>
>>
>>You’re saying that this passes info and relationships? Because they’re 
>>in the same paragraph? It passes in screen readers that can guess the 
>>label of the checkbox? Which ones guess properly?
>>
>>I’m not saying that WCAG requires the code to be written in a specific 
>>way, I’m saying that it requires the relationship association and I 
>>don’t see how a title attribute that duplicates the visible label text 
>>or a checkbox inside the same paragraph as the visible label text 
>>counts as a relationship association.
>>
>>Thank you all for discussing the issue!
>>
>>Paul J. Adam
>>Accessibility Evangelist
>>http://www.deque.com/
>>
>>
>>On Dec 4, 2015, at 3:43 PM, Andrew Kirkpatrick <akirkpat@adobe.com> 
>>wrote:
>>
>>In the instance of a control that is implicitly associated with a 
>>label that may even meet 1.3.1 as well as 4.1.2 through the implicit 
>>means:
>><p><input type=“checkbox”> Please send me a ton of email</p>
>>
>>>On Dec 4, 2015, at 3:43 PM, Andrew Kirkpatrick <akirkpat@adobe.com> 
>>>wrote:
>>>
>>>Does this make sense to you?  Others?
>>
>><PastedGraphic-2.png>
>