Re: Rethinking the necessities of ARIA landmark role "main" and HTML5 <main> element

I have experimented with role="main" in the recent past as well, and it's come up in obscure ways with various clients in the last year as well when trying to diagnose weird accessibility issues  regarding ARIA.

I understand the theory, and I even tried to implement role="main" on WhatSock.com, but the results were not as good as I hoped from what the intended purpose of the role states it to be.

Here is a simple example of this regarding form fields contained within a region with role="main".

<div role="main">

<form>

<input type="text" title="Full name" />

<input type="text" title="Email address" />

</form>

</div>

If you are in Forms Mode in JAWS 13 when using IE, you will here "Landmark Region" in addition to every form field label spoken when pressing Tab, and in JAWS 14 you will hear "Region", which is distracting and quite annoying when dealing with forms that involve more than twenty fields in number for instance.

I've also noticed strange behaviors when other roles are nested within role="main", as well as when interactive widgets are present such as role="tablist", and many others. Some of the behaviors I've seen include the automatic announcement of everything contained within the region of role="main" when dynamic content changes occur within a localized section of another widget also contained within this broad container.

So, after I added role="main" to WhatSock.com, I found all of these issues, and immediately removed it.

Since then, I'm not a fan of adding attributes just because there is a specification that promotes it, without performing comprehensive testing to verify it beforehand.

This is something I recommend everyone do, because I have diagnosed many website issues that are directly a result of developers adding ARIA attributes to their markup without have any idea how it will impact screen reader interaction and feedback.


  ----- Original Message ----- 
  From: Steve Faulkner 
  To: Charles McCathie Nevile 
  Cc: Ian Yang ; w3c-wai-ig@w3.org 
  Sent: Saturday, March 23, 2013 7:14 PM
  Subject: Re: Rethinking the necessities of ARIA landmark role "main" and HTML5 <main> element


  Hi Chaals, 


  thanks for the detailed reply to Ian, the apparent terseness of my own reply was based on the knowledge of Ian's (Yang) being involved in much of the discussion[1] that occurred on the WHATWG list on the subject, and in fact being the person who triggered my renewed interest in the development of the feature. 


  >My conclusions are that differences in the WHAT-WG version are silly, but the element as specified in the HTML specification and as commonly >implemented is actually very useful.


  It should be noted that the differences (with the W3C definition) in how main has been defined in the whatwg spec have not been ignored. I have sought to understand what the reasoning for those differences is [2] and also asked Ian (Hixie) directly on IRC, but have not as yet received any response.


  [1] http://www.w3.org/Search/Mail/Public/search?type-index=public-whatwg-archive&index-type=t&keywords=maincontent&search=Search
  [2] http://lists.w3.org/Archives/Public/public-whatwg-archive/2013Feb/0172.html


  with regards

  --
  SteveF
  HTML 5.1




  On 24 March 2013 00:33, Charles McCathie Nevile <chaals@yandex-team.ru> wrote:

    On Sat, 23 Mar 2013 16:21:17 +0100, Steve Faulkner <faulkner.steve@gmail.com> wrote:


      Hi Ian,


        Ian Hixie, he mentioned that the existence of the ARIA landmark role
        "main" is a mistake



    I've seen this assertion from him, and discussions about why. I am unconvinced by the arguments I have seen. I also haven't seen anything that reasonably contradicts the data Steve produced to justify the element. My discussions with web developers, from small-shop devs to things like Yandex with millions of users across dozens or hundreds of services suggest that the element makes sense.


        That's very thought-provoking


    Well, it might be. The original proposal was thought-provoking enough to also provoke me into reading other people's thoughts and research and even doing a small amount of my own. My conclusions are that differences in the WHAT-WG version are silly, but the element as specified in the HTML specification and as commonly implemented is actually very useful.

    The fact that Ian disagrees with something isn't enough to be though-provoking on its own. He is clever, and often right. But not about everything. Some of his insights into accessibility are very helpful, and some of them just suggest that he knows more about other aspects of HTML.



      both role=main and now <main> are part of the web platform and
      interoperably implemented across browsers and assistive technology



    Yes, and this happened very quickly. That doesn't necessarily mean they are a good idea, because sometimes the wisdom of the crowd isn't quite as clever as we hope, but it suggests that a large proportion of the relevant decision makers, who on balance are usually quite smart and quite thoughtful about what they add to the web, are convinced that the element makes sense.

    A major reason for the element is to replace the "skip to main content" links that are all over the web, for accessibility purposes. While the use of those links is a terrible bit of architecture (they only work if you start from the top of the page and navigate with the keyboard, etc etc) they are deemed useful enough to include on all kinds of websites whose designs have been through multiple rounds of usability testing to ensure they make sense in practice.

    There have been discussions in all kinds of places. Since Steve was the big proponent, he can probably provide pointers by digging through his email archive, but I suggest you look at the mail archives of the W3C's HTML Working Group[1], the W3C's HTML Accessibility Task Force[2] in particular. You can also look at things like IRC logs, blog posts, and so on. A Yandex search [3] shows a handful of interesting perspectives in blogs and articles, too.

    [1] http://lists.w3.org/Archives/Public/public-html/
    [2] http://lists.w3.org/Archives/Public/public-html-a11y/
    [3] http://yandex.com/yandsearch?text=html5+main+element&clid=1823140&lr=213

    Note that this is just my personal opinion, and I am not always right :)

    cheers

    Chaals

    -- 
    Charles McCathie Nevile - Consultant (web standards) CTO Office, Yandex
          chaals@yandex-team.ru         Find more at http://yandex.com