Re: Caption and summary survey

Steven Faulkner 2009-02-20 10.54:
> Hi Josh, thanks for providing clarifications.
> 
>> Some may need configuration (like Window Eyes, but I thought
>> it did work out of the box)
> 
> I tested window eyes version 7 default settings
> 
> When cursoring through content using the arrow keys (browse
> mode) window eyes announces caption and then summary on a data
> table, the same when navigating by table using the T key. I

"Default" vs "out the box" is unclear to me, but I shall put the 
finding online. Did Windows Eyes also place general table info 
inbetween the Caption and the Summary?

> tested JAWS 9 when reading the page, in virtual PC cursor mode,
> the summary is announced before the caption same cursoring
> through the content the content, when naviagting using the T 
> key in virtual PC cursor mode, caption is announced, then
> summary.

I will be putting the finding regarding JAWS 9 online.

> 2009/2/20 Joshue O Connor <joshue.oconnor@cfit.ie>

>> 3) How HTML content is dealt with/announced is primarily a
>> user agent issue. The order in this case is /usually/
>> <caption>, then <summary> but it /could/ also be determined
>> by the order they appear in the source code etc. It is usual
>> to put the <caption> first and the @summary second but the
>> the issue of 'what gets announced when' is peripheral IMO.

Well, I put some more weight to the order, then. I can't make 
sense of the behaviour I experienced in the JAWS 10 demo at all. 
It could also color how (sighted) authors code the pages if they 
experience that summary, in such a significatn UA as JAWS 10, 
behaves more like caption should do.

But for the user, if he always clearly distinguishes between 
@summary and "the rest of the tabble", so that he knows that 
summary is not rendered for anyone other than non-visual user 
agents, then order matters less. My thinking was that it becomes 
easier to discern @summary from caption if it comes after.

It was also very discouraging, I must say, that JAWS 10 treated 
summary "better" than it treated caption. Perhaps someone could 
repeat the test with JAWS 10?

>> 4) A user agent like a screen reader performs other functions
>> like navigation and does not merely /read/ the page. Please
>> note, I have many times come across HTML elements and
>> attributes /not/ being read out or announced when an element
>> is given focus (and I know they are in the code where they
>> should be as I put them there). This can for example be due
>> to issues with a graphics card not playing nice with a
>> version of the screen reader. Anyway, I am making the point
>> that screen readers are not perfect and a quick once off test
>> may not be truly reflective, and I mention this an an aside.

After I, as the last step in my test, installed Dolphin Hal, and 
then deinstalled it again, sound stopped working entirely. But, of 
course, I don't know why Windows Eyes did not tell me the summary. 
I will test it again, if possible. I assumed that I could get it 
to render summary, but I had set my own task to only test the 
out-of-the-box experience.

But I can tell you that Apple VoiceOver has never played it to me. 
It would be interesting to hear from Apple persons if this becuase 
they have not implemented it, or what.

>> 5) You may not be an experienced user of AT. 

I ain't.

>> I am not having
>> a go, but screen readers are very complex. It took me a
>> couple of years of observing and learning from blind friends
>> and colleagues before I became competent and understood many
>> of their quirks. So with all due respect while something may
>> not have worked for you as you expected, it may be other
>> reasons than a fault in the markup as such.

This also why I made the point that I tested without making 
configurations. I tested what I tested. Nothing else. I assume 
that I could have gotten Windows Eyes to read summary (unless it 
was a e.g. soundcard issue). I did not have time to do more in 
dept tests. But I thought it was value in testing out-of-the-box 
experience, which should be somewhat relevant since many authors 
will be testing the same way. But I can put any other experience 
online, if anyone will share with me what they have experienced of 
when I test in more depth.

>> The fact is that complex data tables still need and have a
>> long descriptor in the form of @summary. In my experience
>> FWIW, it is well supported out of the box by Assistive
>> Technology. Some may need configuration (like Window Eyes,
>> but I thought it did work out of the box) in order to
>> announce it but these are primarily /user agent/ issues. The
>> specification needs a long descriptor in the first place.

My test *confirms* that most UAs have some sort of support for 
summary. There were only two UAs that, out of *my* box, did not 
react to summary at all, *in the test I performed*.

>> If you wish to see other useful things you can do. WIth Jaws
>> press the 'T' key and you can jump from table to table with
>> the <caption> and @summaries being announced, very useful.
>> Also you can use INSERT + F3 and bring up the Virtual HTML
>> dialogue which allows you to display the list of any present
>> HTML elements in a page, go to 'Tables' in the and any tables
>> present will be listed in new dialogue box that you can 
>> navigate to using your cursor keys.

As soon as I get sound back, I will test it. ;-) But this was 
alson one limitation I put on myself: The goal was only to kick 
and run.
-- 
leif halvard silli

Received on Friday, 20 February 2009 16:59:18 UTC