E-mail List Archives

Search E-mail List Archives for

Number of posts in this thread: 18 (In chronological order)

From: Lynn Holdsworth
Date: Tue, Dec 30 2014 5:00AM
Subject: Are fieldset and legend still relevant?
No previous message | Next message →

Hi all,

I'm just going through WebAIM's invaluable list of checkpoints, and
found this one under 1.3.1: Info and relationships:

' Text labels are associated with form input elements. Related form
elements are grouped with fieldset/legend.

Rather than using fieldset and legend to group form elements, for a
while now I've been using CSS and ARIA labels.

Fieldsets were always pretty flaky with screenreaders. I have my JAWS
setting set to advanced, and so legends don't get read out as I tab
through the elements inside a fieldset. But ARIA labels do. And they
get read out in the order that the developer deems most sensible.

If I use both legends and ARIA labels, some users must listen to the
labels twice, sometimes once before and once after the info about the
element with focus, which could get pretty confusing for beginners.

I'd love to hear your thoughts on this one.

Cheers, Lynn

From: Birkir R. Gunnarsson
Date: Tue, Dec 30 2014 5:14AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Hi Lynn

It depends how you construct the aria-label I would say.
The relevant pieces of info user needs to ahve access to are the
labels for the individual fields as well as the common label for the
group of fields.
How would you code a gender radiobutton set for instance?
User would want to hear
"gender male"
"gender female"
or at least hear the word "gender" on the first radiobutton, the one
that has focus, right?
Generally I would recommend fieldset with legend "gender" and two
radiobuttons labeled "male" and "female".
If for whatever reason that is not possible I would just construct an
aria group.
<div role="group" aria-labelledby="gender">
<h3 id="gender">Gender</h3>
<label for "gm">Male</label>
<input type="radio" name="rbgender" id="gm">
<label for "gm=f">Female</label>
<input type="radio" name="rbgender" id="gf">
</div>

You could construct an aria-label and add the legend text to one or
more of the radiobuttons.
It would communicate the same info.
The risk with aria-labels in general is that they are invisible and I
often see developers being sloppy and making typos that do not get
caught by their QA teams.

I think the Jaws setting is honestly a bit weird.
Users can be extra super power users with Jaws, but they still cannot
guess what a common legend for a fieldset or a tooltip is on a page
(aria-describedby values are not read by Jaws with verbosity set to
"advanced").
The verbosity setting is not even set for the browser specifically, it
is the global setting.
I think FS should fix this.
I digress. ;)
In short, yes, fieldset/legends are still important to me, though
there are ways one could work around it and code the legend into the
labels.
Cheers

On 12/30/14, Lynn Holdsworth < = EMAIL ADDRESS REMOVED = > wrote:
> Hi all,
>
> I'm just going through WebAIM's invaluable list of checkpoints, and
> found this one under 1.3.1: Info and relationships:
>
> * Text labels are associated with form input elements. Related form
> elements are grouped with fieldset/legend.
>
> Rather than using fieldset and legend to group form elements, for a
> while now I've been using CSS and ARIA labels.
>
> Fieldsets were always pretty flaky with screenreaders. I have my JAWS
> setting set to advanced, and so legends don't get read out as I tab
> through the elements inside a fieldset. But ARIA labels do. And they
> get read out in the order that the developer deems most sensible.
>
> If I use both legends and ARIA labels, some users must listen to the
> labels twice, sometimes once before and once after the info about the
> element with focus, which could get pretty confusing for beginners.
>
> I'd love to hear your thoughts on this one.
>
> Cheers, Lynn
> > > >


--
Work hard. Have fun. Make history.

From: Joy Relton
Date: Tue, Dec 30 2014 6:01AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Hi all,

As a JAWS user myself, I suggest the following for consideration

Pages and code should be written and tested with beginner settings, the goal
is to code so that persons with disabilities can access all of the
information, not to test the level of proficiency in the use of assistive
technology.

The conformance with guidelines or standards should not be measured by the
performance of one assistive technology. Since I am writing this one-handed
due to the temporary disability of one hand being sprained, I am envious of
those with speech recognition. Unless things have chaned significantly, a
user of speech recognition cannot access and use forms which do not convey
both the information conveyed by the fieldset and the legend.

From: Lynn Holdsworth
Date: Tue, Dec 30 2014 6:13AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Hi Birkir,

Sorry, I should have made myself clearer. I'm not advocating using
invisible labels, but rather using ARIA-LABELLEDBY to associate a form
field with a piece of text that's visible somewhere appropriate on the
page.

I've knocked up some examples. With JAWS 15 set to advanced verbosity,
the second and third examples work beautifully, while in the first the
"Gender" label is ignored.

http://behindtheyellowline.org.uk/labelledby.htm

(Excuse any validation errors - quick turn-around time and all that!)

I'll try these with NVDA and VoiceOver when I get a moment, and may
create some more complex examples to illustrate what I'm trying to
achieve.

But for now, would fieldset and legend alone cover WCAG2?

Thanks, Lynn

On 30/12/2014, Birkir R. Gunnarsson < = EMAIL ADDRESS REMOVED = > wrote:
> Hi Lynn
>
> It depends how you construct the aria-label I would say.
> The relevant pieces of info user needs to ahve access to are the
> labels for the individual fields as well as the common label for the
> group of fields.
> How would you code a gender radiobutton set for instance?
> User would want to hear
> "gender male"
> "gender female"
> or at least hear the word "gender" on the first radiobutton, the one
> that has focus, right?
> Generally I would recommend fieldset with legend "gender" and two
> radiobuttons labeled "male" and "female".
> If for whatever reason that is not possible I would just construct an
> aria group.
> <div role="group" aria-labelledby="gender">
> <h3 id="gender">Gender</h3>
> <label for "gm">Male</label>
> <input type="radio" name="rbgender" id="gm">
> <label for "gm=f">Female</label>
> <input type="radio" name="rbgender" id="gf">
> </div>
>
> You could construct an aria-label and add the legend text to one or
> more of the radiobuttons.
> It would communicate the same info.
> The risk with aria-labels in general is that they are invisible and I
> often see developers being sloppy and making typos that do not get
> caught by their QA teams.
>
> I think the Jaws setting is honestly a bit weird.
> Users can be extra super power users with Jaws, but they still cannot
> guess what a common legend for a fieldset or a tooltip is on a page
> (aria-describedby values are not read by Jaws with verbosity set to
> "advanced").
> The verbosity setting is not even set for the browser specifically, it
> is the global setting.
> I think FS should fix this.
> I digress. ;)
> In short, yes, fieldset/legends are still important to me, though
> there are ways one could work around it and code the legend into the
> labels.
> Cheers
>
> On 12/30/14, Lynn Holdsworth < = EMAIL ADDRESS REMOVED = > wrote:
>> Hi all,
>>
>> I'm just going through WebAIM's invaluable list of checkpoints, and
>> found this one under 1.3.1: Info and relationships:
>>
>> * Text labels are associated with form input elements. Related form
>> elements are grouped with fieldset/legend.
>>
>> Rather than using fieldset and legend to group form elements, for a
>> while now I've been using CSS and ARIA labels.
>>
>> Fieldsets were always pretty flaky with screenreaders. I have my JAWS
>> setting set to advanced, and so legends don't get read out as I tab
>> through the elements inside a fieldset. But ARIA labels do. And they
>> get read out in the order that the developer deems most sensible.
>>
>> If I use both legends and ARIA labels, some users must listen to the
>> labels twice, sometimes once before and once after the info about the
>> element with focus, which could get pretty confusing for beginners.
>>
>> I'd love to hear your thoughts on this one.
>>
>> Cheers, Lynn
>> >> >> >>
>
>
> --
> Work hard. Have fun. Make history.
> > > >

From: Lynn Holdsworth
Date: Tue, Dec 30 2014 6:20AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Hi Joy,

Thanks for your comments and the good points you make.

I believe the user is King and we should do everything possible to get
as close to total inclusivity as we can.

I take your point that it's frustrating to have to code for one group
of users whose AT is set to a particular setting. But I would guess
that a significant enough number of people are using JAWS with
intermediate or advanced verbosity settings to make that bit of extra
effort worthwhile.

Does the WebAIM survey collect data on verbosity settings? If not it
might be worth considering this for next year's survey.

Thanks, Lynn

On 30/12/2014, Lynn Holdsworth < = EMAIL ADDRESS REMOVED = > wrote:
> Hi Birkir,
>
> Sorry, I should have made myself clearer. I'm not advocating using
> invisible labels, but rather using ARIA-LABELLEDBY to associate a form
> field with a piece of text that's visible somewhere appropriate on the
> page.
>
> I've knocked up some examples. With JAWS 15 set to advanced verbosity,
> the second and third examples work beautifully, while in the first the
> "Gender" label is ignored.
>
> http://behindtheyellowline.org.uk/labelledby.htm
>
> (Excuse any validation errors - quick turn-around time and all that!)
>
> I'll try these with NVDA and VoiceOver when I get a moment, and may
> create some more complex examples to illustrate what I'm trying to
> achieve.
>
> But for now, would fieldset and legend alone cover WCAG2?
>
> Thanks, Lynn
>
> On 30/12/2014, Birkir R. Gunnarsson < = EMAIL ADDRESS REMOVED = > wrote:
>> Hi Lynn
>>
>> It depends how you construct the aria-label I would say.
>> The relevant pieces of info user needs to ahve access to are the
>> labels for the individual fields as well as the common label for the
>> group of fields.
>> How would you code a gender radiobutton set for instance?
>> User would want to hear
>> "gender male"
>> "gender female"
>> or at least hear the word "gender" on the first radiobutton, the one
>> that has focus, right?
>> Generally I would recommend fieldset with legend "gender" and two
>> radiobuttons labeled "male" and "female".
>> If for whatever reason that is not possible I would just construct an
>> aria group.
>> <div role="group" aria-labelledby="gender">
>> <h3 id="gender">Gender</h3>
>> <label for "gm">Male</label>
>> <input type="radio" name="rbgender" id="gm">
>> <label for "gm=f">Female</label>
>> <input type="radio" name="rbgender" id="gf">
>> </div>
>>
>> You could construct an aria-label and add the legend text to one or
>> more of the radiobuttons.
>> It would communicate the same info.
>> The risk with aria-labels in general is that they are invisible and I
>> often see developers being sloppy and making typos that do not get
>> caught by their QA teams.
>>
>> I think the Jaws setting is honestly a bit weird.
>> Users can be extra super power users with Jaws, but they still cannot
>> guess what a common legend for a fieldset or a tooltip is on a page
>> (aria-describedby values are not read by Jaws with verbosity set to
>> "advanced").
>> The verbosity setting is not even set for the browser specifically, it
>> is the global setting.
>> I think FS should fix this.
>> I digress. ;)
>> In short, yes, fieldset/legends are still important to me, though
>> there are ways one could work around it and code the legend into the
>> labels.
>> Cheers
>>
>> On 12/30/14, Lynn Holdsworth < = EMAIL ADDRESS REMOVED = > wrote:
>>> Hi all,
>>>
>>> I'm just going through WebAIM's invaluable list of checkpoints, and
>>> found this one under 1.3.1: Info and relationships:
>>>
>>> * Text labels are associated with form input elements. Related form
>>> elements are grouped with fieldset/legend.
>>>
>>> Rather than using fieldset and legend to group form elements, for a
>>> while now I've been using CSS and ARIA labels.
>>>
>>> Fieldsets were always pretty flaky with screenreaders. I have my JAWS
>>> setting set to advanced, and so legends don't get read out as I tab
>>> through the elements inside a fieldset. But ARIA labels do. And they
>>> get read out in the order that the developer deems most sensible.
>>>
>>> If I use both legends and ARIA labels, some users must listen to the
>>> labels twice, sometimes once before and once after the info about the
>>> element with focus, which could get pretty confusing for beginners.
>>>
>>> I'd love to hear your thoughts on this one.
>>>
>>> Cheers, Lynn
>>> >>> >>> >>>
>>
>>
>> --
>> Work hard. Have fun. Make history.
>> >> >> >>
>

From: Thomas McKeithan II
Date: Tue, Dec 30 2014 8:12AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

I concur with Joy. We should determine conformance to the standards basd upon how the page is coded and not how a specific AT Tool behavors. For years I've advocated fot testing using AT Tools with the "out of the Box" settings rather than customizations.

Respectfully,
Thomas Lee McKeithan II
QSSI
http://www.qssinc.com
508 SME, SSQA Solutions Center
10480 Little Patuxent Pkwy , Suite 350
Columbia , MD 21044
(301 )977-7884 x1058 (Work)
(202) 276-6437 (Cell)
 

This electronic mail (including any attachments) may contain information that is privileged, confidential, and/or otherwise protected from disclosure to anyone other than its intended recipient(s). Any dissemination or use of this electronic email or its contents (including any attachments) by persons other than the intended recipient(s) is strictly prohibited. If you have received this message in error, please notify the sender by reply email and delete the original message (including any attachments) in its entirety.

From: Lynn Holdsworth
Date: Tue, Dec 30 2014 8:34AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Hi Thomas,

I guess I've been conflating testing with development. I'm both a
tester and dev, and I tend to do that. Apologies for any confusion.

If something works with most AT in its out-of-the-box state, but not
with one slightly tweaked popular AT, I don't think it would do any
harm to recommend to developers a few changes that would make it work
for more people.

Kind regards, Lynn

On 30/12/2014, Thomas McKeithan II < = EMAIL ADDRESS REMOVED = > wrote:
> I concur with Joy. We should determine conformance to the standards basd
> upon how the page is coded and not how a specific AT Tool behavors. For
> years I've advocated fot testing using AT Tools with the "out of the Box"
> settings rather than customizations.
>
> Respectfully,
> Thomas Lee McKeithan II
> QSSI
> http://www.qssinc.com
> 508 SME, SSQA Solutions Center
> 10480 Little Patuxent Pkwy , Suite 350
> Columbia , MD 21044
> (301 )977-7884 x1058 (Work)
> (202) 276-6437 (Cell)
>
>
> This electronic mail (including any attachments) may contain information
> that is privileged, confidential, and/or otherwise protected from disclosure
> to anyone other than its intended recipient(s). Any dissemination or use of
> this electronic email or its contents (including any attachments) by persons
> other than the intended recipient(s) is strictly prohibited. If you have
> received this message in error, please notify the sender by reply email and
> delete the original message (including any attachments) in its entirety.
>
>
>

From: Birkir R. Gunnarsson
Date: Tue, Dec 30 2014 8:40AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Hey guys

Joy, well said.
For the examples, at a glance.
Yes, the labeling in examples 2 and 3 works, and I would pass it.
However it has three drawbacks.
- Using aria-labelledby rather than the html labeling decreases the
clickable area of the radiobuttons (for users with dexterity
impairments larger clickable area is always a plus). When using label
for user can click on the label and focus moves into its associated
radiobutton. With aria-labeling this does not happen.

- Fielset/legend construct usually draws a visual box around the
legend and choices, which is good for users with cognitive
impairments. using ARIA group role does not do this, nor the aria
labeling, .. though the page designer can always create these types of
things using CSS anyway.

- this set up will force screen readers to read legend for every
single radiobutton choice, which can get very annoying when the legend
is very wrong.
Using fieldset/legend or corresponding ARIA group role users can
customize, at least to some degree, how often they want to hear the
legend.
I only want to hear it once, when focus moves into the radiobuttons.
I have not looked at Survey Monkey lately, but it used to drive me
bananas, apples and pairs, to try and take their surveys, because the
bloody question (and instructions sometimes) was read out in its
entirety before the label of any radiobutton choice.
Again, I do not know what they used there, I would have to go back,
check, and hope they have replaced whatever structure they used.


So, again, your way is not wrong, I would not fail it, but it has some
drawbacks that I would point out if I were doing an accessibility
assessment of the page, at least best practice items to discuss, not
outright violations.
Note though that I only took a quick look, it is one of those crazy
days at work, so I have not had time to analyze the code in any depth.

Cheers
-B

On 12/30/14, Thomas McKeithan II < = EMAIL ADDRESS REMOVED = > wrote:
> I concur with Joy. We should determine conformance to the standards basd
> upon how the page is coded and not how a specific AT Tool behavors. For
> years I've advocated fot testing using AT Tools with the "out of the Box"
> settings rather than customizations.
>
> Respectfully,
> Thomas Lee McKeithan II
> QSSI
> http://www.qssinc.com
> 508 SME, SSQA Solutions Center
> 10480 Little Patuxent Pkwy , Suite 350
> Columbia , MD 21044
> (301 )977-7884 x1058 (Work)
> (202) 276-6437 (Cell)
>
>
> This electronic mail (including any attachments) may contain information
> that is privileged, confidential, and/or otherwise protected from disclosure
> to anyone other than its intended recipient(s). Any dissemination or use of
> this electronic email or its contents (including any attachments) by persons
> other than the intended recipient(s) is strictly prohibited. If you have
> received this message in error, please notify the sender by reply email and
> delete the original message (including any attachments) in its entirety.
>
>
>

From: Lynn Holdsworth
Date: Tue, Dec 30 2014 8:45AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

http://behindtheyellowline.org.uk/labelledby.htm

OK, have just tested these three examples with VoiceOver on iOS 8 in
its default state, and got similar results, except that using
VoiceOver I can't seem to see a fieldset's legend at all.

So I think I'll stick my neck out and propose that now we have
solutions other than using fieldset and legend, the fieldset/legend
solution by itself isn't enough.

Option 3, using a fieldset's legend as an ARIA label (thank you Birkir
for this idea), seems to cover all bases.

I'd really appreciate any thoughts on this.

Kind regards, Lynn

On 30/12/2014, Lynn Holdsworth < = EMAIL ADDRESS REMOVED = > wrote:
> Hi Thomas,
>
> I guess I've been conflating testing with development. I'm both a
> tester and dev, and I tend to do that. Apologies for any confusion.
>
> If something works with most AT in its out-of-the-box state, but not
> with one slightly tweaked popular AT, I don't think it would do any
> harm to recommend to developers a few changes that would make it work
> for more people.
>
> Kind regards, Lynn
>
> On 30/12/2014, Thomas McKeithan II < = EMAIL ADDRESS REMOVED = > wrote:
>> I concur with Joy. We should determine conformance to the standards basd
>> upon how the page is coded and not how a specific AT Tool behavors. For
>> years I've advocated fot testing using AT Tools with the "out of the Box"
>> settings rather than customizations.
>>
>> Respectfully,
>> Thomas Lee McKeithan II
>> QSSI
>> http://www.qssinc.com
>> 508 SME, SSQA Solutions Center
>> 10480 Little Patuxent Pkwy , Suite 350
>> Columbia , MD 21044
>> (301 )977-7884 x1058 (Work)
>> (202) 276-6437 (Cell)
>>
>>
>> This electronic mail (including any attachments) may contain information
>> that is privileged, confidential, and/or otherwise protected from
>> disclosure
>> to anyone other than its intended recipient(s). Any dissemination or use
>> of
>> this electronic email or its contents (including any attachments) by
>> persons
>> other than the intended recipient(s) is strictly prohibited. If you have
>> received this message in error, please notify the sender by reply email
>> and
>> delete the original message (including any attachments) in its entirety.
>>
>>
>>

From: Lynn Holdsworth
Date: Tue, Dec 30 2014 8:55AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Hi Birkir,

Example 3 has a fieldset/legend, some ARIA labels and an explicitly
associated label. So you still get the box drawn around the group as
well as the clickable labels.

As regards the third drawback, I take your point there. Even if the
labels are ordered in a way that causes the most important one to be
spoken first, one must still wait until all labels are read out before
being able to hear the type and state of the control. It's possible to
customise JAWS and possibly other screenreaders to speak control type
and state first, but most users are unlikely to know how to do this.

Using a fieldset and legend seems to have the same drawback. I can't
find a way using VoiceOver to read the legend, nor in JAWS to only
read it once when an item in the group receives focus.

So on balance I'd still propose that the combo in example 3 is the
most comprehensive solution.

Kind regards, Lynn

On 30/12/2014, Birkir R. Gunnarsson < = EMAIL ADDRESS REMOVED = > wrote:
> Hey guys
>
> Joy, well said.
> For the examples, at a glance.
> Yes, the labeling in examples 2 and 3 works, and I would pass it.
> However it has three drawbacks.
> - Using aria-labelledby rather than the html labeling decreases the
> clickable area of the radiobuttons (for users with dexterity
> impairments larger clickable area is always a plus). When using label
> for user can click on the label and focus moves into its associated
> radiobutton. With aria-labeling this does not happen.
>
> - Fielset/legend construct usually draws a visual box around the
> legend and choices, which is good for users with cognitive
> impairments. using ARIA group role does not do this, nor the aria
> labeling, .. though the page designer can always create these types of
> things using CSS anyway.
>
> - this set up will force screen readers to read legend for every
> single radiobutton choice, which can get very annoying when the legend
> is very wrong.
> Using fieldset/legend or corresponding ARIA group role users can
> customize, at least to some degree, how often they want to hear the
> legend.
> I only want to hear it once, when focus moves into the radiobuttons.
> I have not looked at Survey Monkey lately, but it used to drive me
> bananas, apples and pairs, to try and take their surveys, because the
> bloody question (and instructions sometimes) was read out in its
> entirety before the label of any radiobutton choice.
> Again, I do not know what they used there, I would have to go back,
> check, and hope they have replaced whatever structure they used.
>
>
> So, again, your way is not wrong, I would not fail it, but it has some
> drawbacks that I would point out if I were doing an accessibility
> assessment of the page, at least best practice items to discuss, not
> outright violations.
> Note though that I only took a quick look, it is one of those crazy
> days at work, so I have not had time to analyze the code in any depth.
>
> Cheers
> -B
>
> On 12/30/14, Thomas McKeithan II < = EMAIL ADDRESS REMOVED = > wrote:
>> I concur with Joy. We should determine conformance to the standards basd
>> upon how the page is coded and not how a specific AT Tool behavors. For
>> years I've advocated fot testing using AT Tools with the "out of the Box"
>> settings rather than customizations.
>>
>> Respectfully,
>> Thomas Lee McKeithan II
>> QSSI
>> http://www.qssinc.com
>> 508 SME, SSQA Solutions Center
>> 10480 Little Patuxent Pkwy , Suite 350
>> Columbia , MD 21044
>> (301 )977-7884 x1058 (Work)
>> (202) 276-6437 (Cell)
>>
>>
>> This electronic mail (including any attachments) may contain information
>> that is privileged, confidential, and/or otherwise protected from
>> disclosure
>> to anyone other than its intended recipient(s). Any dissemination or use
>> of
>> this electronic email or its contents (including any attachments) by
>> persons
>> other than the intended recipient(s) is strictly prohibited. If you have
>> received this message in error, please notify the sender by reply email
>> and
>> delete the original message (including any attachments) in its entirety.
>>
>>
>>

From: Birkir R. Gunnarsson
Date: Tue, Dec 30 2014 9:04AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Hey Lynn

Absolutely, well spoken, the third example does look the best, and
having the most appropriate label first allows you to quickly skip
what you are not interested in rather than listening to it.

Honestly what we need to do more of, as a group, is to actively
request fixes in our assistive technologies, rather than forcing
ourselves and others to keep coding around individual bugs or
questionable interpretations.

But in the meantime we need to code pages in such a way as to be
accessible to the largest group of users, at least when we have the
luxury of doing so.
We can't require people to do so .. i.e. if their code is compliant
with the standard, they may simply not have time, resources or
interest to make the perfect page, but we always celebrate it when
people can do that, and we definitely do well to at least propose how
it can be done.
Cheers and thanks for the examples, I will look at the page in more
detail later.


On 12/30/14, Lynn Holdsworth < = EMAIL ADDRESS REMOVED = > wrote:
> Hi Birkir,
>
> Example 3 has a fieldset/legend, some ARIA labels and an explicitly
> associated label. So you still get the box drawn around the group as
> well as the clickable labels.
>
> As regards the third drawback, I take your point there. Even if the
> labels are ordered in a way that causes the most important one to be
> spoken first, one must still wait until all labels are read out before
> being able to hear the type and state of the control. It's possible to
> customise JAWS and possibly other screenreaders to speak control type
> and state first, but most users are unlikely to know how to do this.
>
> Using a fieldset and legend seems to have the same drawback. I can't
> find a way using VoiceOver to read the legend, nor in JAWS to only
> read it once when an item in the group receives focus.
>
> So on balance I'd still propose that the combo in example 3 is the
> most comprehensive solution.
>
> Kind regards, Lynn
>
> On 30/12/2014, Birkir R. Gunnarsson < = EMAIL ADDRESS REMOVED = > wrote:
>> Hey guys
>>
>> Joy, well said.
>> For the examples, at a glance.
>> Yes, the labeling in examples 2 and 3 works, and I would pass it.
>> However it has three drawbacks.
>> - Using aria-labelledby rather than the html labeling decreases the
>> clickable area of the radiobuttons (for users with dexterity
>> impairments larger clickable area is always a plus). When using label
>> for user can click on the label and focus moves into its associated
>> radiobutton. With aria-labeling this does not happen.
>>
>> - Fielset/legend construct usually draws a visual box around the
>> legend and choices, which is good for users with cognitive
>> impairments. using ARIA group role does not do this, nor the aria
>> labeling, .. though the page designer can always create these types of
>> things using CSS anyway.
>>
>> - this set up will force screen readers to read legend for every
>> single radiobutton choice, which can get very annoying when the legend
>> is very wrong.
>> Using fieldset/legend or corresponding ARIA group role users can
>> customize, at least to some degree, how often they want to hear the
>> legend.
>> I only want to hear it once, when focus moves into the radiobuttons.
>> I have not looked at Survey Monkey lately, but it used to drive me
>> bananas, apples and pairs, to try and take their surveys, because the
>> bloody question (and instructions sometimes) was read out in its
>> entirety before the label of any radiobutton choice.
>> Again, I do not know what they used there, I would have to go back,
>> check, and hope they have replaced whatever structure they used.
>>
>>
>> So, again, your way is not wrong, I would not fail it, but it has some
>> drawbacks that I would point out if I were doing an accessibility
>> assessment of the page, at least best practice items to discuss, not
>> outright violations.
>> Note though that I only took a quick look, it is one of those crazy
>> days at work, so I have not had time to analyze the code in any depth.
>>
>> Cheers
>> -B
>>
>> On 12/30/14, Thomas McKeithan II < = EMAIL ADDRESS REMOVED = > wrote:
>>> I concur with Joy. We should determine conformance to the standards basd
>>> upon how the page is coded and not how a specific AT Tool behavors. For
>>> years I've advocated fot testing using AT Tools with the "out of the
>>> Box"
>>> settings rather than customizations.
>>>
>>> Respectfully,
>>> Thomas Lee McKeithan II
>>> QSSI
>>> http://www.qssinc.com
>>> 508 SME, SSQA Solutions Center
>>> 10480 Little Patuxent Pkwy , Suite 350
>>> Columbia , MD 21044
>>> (301 )977-7884 x1058 (Work)
>>> (202) 276-6437 (Cell)
>>>
>>>
>>> This electronic mail (including any attachments) may contain information
>>> that is privileged, confidential, and/or otherwise protected from
>>> disclosure
>>> to anyone other than its intended recipient(s). Any dissemination or use
>>> of
>>> this electronic email or its contents (including any attachments) by
>>> persons
>>> other than the intended recipient(s) is strictly prohibited. If you have
>>> received this message in error, please notify the sender by reply email
>>> and
>>> delete the original message (including any attachments) in its entirety.
>>>
>>>
>>>

From: Jared Smith
Date: Tue, Dec 30 2014 9:44AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Birkir R. Gunnarsson wrote:

> Honestly what we need to do more of, as a group, is to actively
> request fixes in our assistive technologies, rather than forcing
> ourselves and others to keep coding around individual bugs or
> questionable interpretations.

This! It is inexplicable that VoiceOver and (at last check)
Window-Eyes do not yet support reading legend when navigating to
controls within a fieldset. It also doesn't make sense that JAWS would
ever ignore this critical association, even when the verbosity
settings are cranked up.

I certainly would not consider using fieldset/legend alone a WCAG
failure. It is, however, clearly an ATAG failure if a screen reader
doesn't support them.

As far as ARIA approaches to address these deficiencies, for radio
buttons, a proper approach would be to use role="radiogroup" on the
grouping, and aria-labelledby to reference the 'legend'. But alas,
this also is not yet supported in VoiceOver.

Jared

From: Paul J. Adam
Date: Tue, Dec 30 2014 10:15AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Data table accessibility is also broken in iOS 8 and VoiceOver, it worked in 7. Fieldset/Legend has never worked on iOS but used to work on OS X, now it's broken in OS X. Data table accessibility does work in OS X thankfully!

Firefox on Android with TalkBack has the best mobile accessibility support! It handles Fieldset/Legend perfectly by reading the Legend once when setting focus to an input in the Fieldset whether from the top or bottom of the Fieldset.

It sucks when you recommend using Fieldset/Legend and then VoiceOver never works with it! I also don't want to recommend slapping aria-labelledby on every single input when you could just use HTML4 semantics.


Paul J. Adam
Accessibility Evangelist
www.deque.com <http://www.deque.com/>;
> On Dec 30, 2014, at 10:44 AM, Jared Smith < = EMAIL ADDRESS REMOVED = > wrote:
>
> Birkir R. Gunnarsson wrote:
>
>> Honestly what we need to do more of, as a group, is to actively
>> request fixes in our assistive technologies, rather than forcing
>> ourselves and others to keep coding around individual bugs or
>> questionable interpretations.
>
> This! It is inexplicable that VoiceOver and (at last check)
> Window-Eyes do not yet support reading legend when navigating to
> controls within a fieldset. It also doesn't make sense that JAWS would
> ever ignore this critical association, even when the verbosity
> settings are cranked up.
>
> I certainly would not consider using fieldset/legend alone a WCAG
> failure. It is, however, clearly an ATAG failure if a screen reader
> doesn't support them.
>
> As far as ARIA approaches to address these deficiencies, for radio
> buttons, a proper approach would be to use role="radiogroup" on the
> grouping, and aria-labelledby to reference the 'legend'. But alas,
> this also is not yet supported in VoiceOver.
>
> Jared
> > >

From: Birkir R. Gunnarsson
Date: Tue, Dec 30 2014 11:25AM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Jaws will not announce aria-checked attribute when using
radio/radiogroup in IE (10 and 11).
Works in FF.
It is frustrating when we try to recommend the right things to do, but
users end up equally unhappy and keep complaining, because the
technologies they are working with do not adhere to the standards.
I will have to give a thumbs up to NVDA who have been both good at
complying and being responsive to issues when they have failed to
comply or do an odd job of it.
I am concerned that Voiceover appears to be moving in the wrong
direction when it comes to supporting web content.
I still just absolutely fail to grasp how a screen reader announcing
"collapsed" for aria-expanded="true", and expanded for
aria-expanded="false" (Voiceover on iOS 8.1 .. even after 2 updates)
gets through QA testing.
Ok, end of screen reader rant, it is off-topic, but one has to vent a
little sometimes *grin*.
I seriously hope that Voiceover for iOS, leader and inventor of the
accessible touch screen, stops moving in the wrong direction for web
content.
Cheers
-B

On 12/30/14, Paul J. Adam < = EMAIL ADDRESS REMOVED = > wrote:
> Data table accessibility is also broken in iOS 8 and VoiceOver, it worked in
> 7. Fieldset/Legend has never worked on iOS but used to work on OS X, now
> it's broken in OS X. Data table accessibility does work in OS X thankfully!
>
> Firefox on Android with TalkBack has the best mobile accessibility support!
> It handles Fieldset/Legend perfectly by reading the Legend once when setting
> focus to an input in the Fieldset whether from the top or bottom of the
> Fieldset.
>
> It sucks when you recommend using Fieldset/Legend and then VoiceOver never
> works with it! I also don't want to recommend slapping aria-labelledby on
> every single input when you could just use HTML4 semantics.
>
>
> Paul J. Adam
> Accessibility Evangelist
> www.deque.com <http://www.deque.com/>;
>> On Dec 30, 2014, at 10:44 AM, Jared Smith < = EMAIL ADDRESS REMOVED = > wrote:
>>
>> Birkir R. Gunnarsson wrote:
>>
>>> Honestly what we need to do more of, as a group, is to actively
>>> request fixes in our assistive technologies, rather than forcing
>>> ourselves and others to keep coding around individual bugs or
>>> questionable interpretations.
>>
>> This! It is inexplicable that VoiceOver and (at last check)
>> Window-Eyes do not yet support reading legend when navigating to
>> controls within a fieldset. It also doesn't make sense that JAWS would
>> ever ignore this critical association, even when the verbosity
>> settings are cranked up.
>>
>> I certainly would not consider using fieldset/legend alone a WCAG
>> failure. It is, however, clearly an ATAG failure if a screen reader
>> doesn't support them.
>>
>> As far as ARIA approaches to address these deficiencies, for radio
>> buttons, a proper approach would be to use role="radiogroup" on the
>> grouping, and aria-labelledby to reference the 'legend'. But alas,
>> this also is not yet supported in VoiceOver.
>>
>> Jared
>> >> >> >
> > > >


--
Work hard. Have fun. Make history.

From: Joy Relton
Date: Tue, Dec 30 2014 12:38PM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

Lynn,

I agree with you. The AT companies often do a great job, when they miss the mark or things don't work we should bring this to their attention. I'm glad to see that fewer AT vendors are attempting to solve the problems by scripting. In my opinion, scriptin should be the exception, not the rule for making technology accessible. As Thomas said, test to the nstandards and guidelines.

From: Thomas McKeithan II
Date: Tue, Dec 30 2014 12:50PM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

True.

Respectfully,
Thomas Lee McKeithan II
QSSI
http://www.qssinc.com
508 SME, SSQA Solutions Center
10480 Little Patuxent Pkwy , Suite 350
Columbia , MD 21044
(301 )977-7884 x1058 (Work)
(202) 276-6437 (Cell)
 

This electronic mail (including any attachments) may contain information that is privileged, confidential, and/or otherwise protected from disclosure to anyone other than its intended recipient(s). Any dissemination or use of this electronic email or its contents (including any attachments) by persons other than the intended recipient(s) is strictly prohibited. If you have received this message in error, please notify the sender by reply email and delete the original message (including any attachments) in its entirety.

From: Jonathan Avila
Date: Tue, Dec 30 2014 7:52PM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | Next message →

> by scripting. In my opinion, scriptin should be the exception, not the rule for making technology accessible. As Thomas said, test to the nstandards and guidelines.

While I agree that we shouldn't be measuring against specific ATs -- techniques have to be practical -- that is, they have to be accessibility supported. This assertion is in place to prevent developers from creating some framework that exposes information or makes a claim to expose information in programmatic ways but the ways are not practically usable by AT. For example, I could claim that divs on my webpage use the title attribute to expose information to screen reader users. In reality a title attribute on a div is not reasonable way of exposing information at this time so the community has determined that a technique such as this is not a sufficient technique.

Jonathan

From: Jonathan C. Cohn
Date: Thu, Jan 01 2015 4:10PM
Subject: Re: Are fieldset and legend still relevant?
← Previous message | No next message

I finally got a chance to look at the examples you provided. In VoiceOver, the FieldSet appears as a Group box in both DOM and Group mode. One needs to interact/ stop interact with the FieldSet in order to change the radio buttons, so when one sees gender while navigating wit the VoiceOver cursor neither example one or example 2 will tell you what gender is selected until you interact with the FieldSet and examine the selected radio button. and if one clicks on goto Element commands it will first go to the Gender FieldSet box and then into the FieldSet to the Radio Buttons. So While tabbing will not read the gender information at least not in default verbosity, it does recognize that the FieldSet is a grouping and requires the VoiceOVer user to interact with the field set to edit any changes. I have actually found that this is a bit annoying in the VoiceOVer web implementation .



In JAWs 15, Advanced mode does not speak Group Box names, and since at least VoiceOver considers the FieldSet a group Box with the Legend being the GroupBox Name, it is not terribly surprising that advanced mode in JAWS does not speak the FieldSet legend. In any case when using the insert-tab I believe JAWS will speak the Group Box name even if it is not enabled to be spoken with default navigation. I therefore am declaring that JAWS not reading the FFieldSet in Advanced mode is a design decision and not a bug.

Jonathan Cohn
to the “Mail” radio button.
Best wishes,

Jonathan



On Dec 30, 2014, at 10:55, Lynn Holdsworth < = EMAIL ADDRESS REMOVED = > wrote:

> Hi Birkir,
>
> Example 3 has a fieldset/legend, some ARIA labels and an explicitly
> associated label. So you still get the box drawn around the group as
> well as the clickable labels.
>
> As regards the third drawback, I take your point there. Even if the
> labels are ordered in a way that causes the most important one to be
> spoken first, one must still wait until all labels are read out before
> being able to hear the type and state of the control. It's possible to
> customise JAWS and possibly other screenreaders to speak control type
> and state first, but most users are unlikely to know how to do this.
>
> Using a fieldset and legend seems to have the same drawback. I can't
> find a way using VoiceOver to read the legend, nor in JAWS to only
> read it once when an item in the group receives focus.
>
> So on balance I'd still propose that the combo in example 3 is the
> most comprehensive solution.
>
> Kind regards, Lynn
>
> On 30/12/2014, Birkir R. Gunnarsson < = EMAIL ADDRESS REMOVED = > wrote:
>> Hey guys
>>
>> Joy, well said.
>> For the examples, at a glance.
>> Yes, the labeling in examples 2 and 3 works, and I would pass it.
>> However it has three drawbacks.
>> - Using aria-labelledby rather than the html labeling decreases the
>> clickable area of the radiobuttons (for users with dexterity
>> impairments larger clickable area is always a plus). When using label
>> for user can click on the label and focus moves into its associated
>> radiobutton. With aria-labeling this does not happen.
>>
>> - Fielset/legend construct usually draws a visual box around the
>> legend and choices, which is good for users with cognitive
>> impairments. using ARIA group role does not do this, nor the aria
>> labeling, .. though the page designer can always create these types of
>> things using CSS anyway.
>>
>> - this set up will force screen readers to read legend for every
>> single radiobutton choice, which can get very annoying when the legend
>> is very wrong.
>> Using fieldset/legend or corresponding ARIA group role users can
>> customize, at least to some degree, how often they want to hear the
>> legend.
>> I only want to hear it once, when focus moves into the radiobuttons.
>> I have not looked at Survey Monkey lately, but it used to drive me
>> bananas, apples and pairs, to try and take their surveys, because the
>> bloody question (and instructions sometimes) was read out in its
>> entirety before the label of any radiobutton choice.
>> Again, I do not know what they used there, I would have to go back,
>> check, and hope they have replaced whatever structure they used.
>>
>>
>> So, again, your way is not wrong, I would not fail it, but it has some
>> drawbacks that I would point out if I were doing an accessibility
>> assessment of the page, at least best practice items to discuss, not
>> outright violations.
>> Note though that I only took a quick look, it is one of those crazy
>> days at work, so I have not had time to analyze the code in any depth.
>>
>> Cheers
>> -B
>>
>> On 12/30/14, Thomas McKeithan II < = EMAIL ADDRESS REMOVED = > wrote:
>>> I concur with Joy. We should determine conformance to the standards basd
>>> upon how the page is coded and not how a specific AT Tool behavors. For
>>> years I've advocated fot testing using AT Tools with the "out of the Box"
>>> settings rather than customizations.
>>>
>>> Respectfully,
>>> Thomas Lee McKeithan II
>>> QSSI
>>> http://www.qssinc.com
>>> 508 SME, SSQA Solutions Center
>>> 10480 Little Patuxent Pkwy , Suite 350
>>> Columbia , MD 21044
>>> (301 )977-7884 x1058 (Work)
>>> (202) 276-6437 (Cell)
>>>
>>>
>>> This electronic mail (including any attachments) may contain information
>>> that is privileged, confidential, and/or otherwise protected from
>>> disclosure
>>> to anyone other than its intended recipient(s). Any dissemination or use
>>> of
>>> this electronic email or its contents (including any attachments) by
>>> persons
>>> other than the intended recipient(s) is strictly prohibited. If you have
>>> received this message in error, please notify the sender by reply email
>>> and
>>> delete the original message (including any attachments) in its entirety.
>>>
>>>
>>>