- From: James Craig <jcraig@apple.com>
- Date: Wed, 19 Nov 2008 12:22:57 -0800
- To: Chris Blouch <cblouch@aol.com>
- Cc: David Bolter <david.bolter@utoronto.ca>, Joseph Scheuhammer <clown@utoronto.ca>, "wai-xtech@w3.org" <wai-xtech@w3.org>, earl johnson <earlj.biker@gmail.com>
Chris Blouch wrote: > I spent quite a bit of time trying to implement keyboard shortcuts > that didn't interfere with either the OS, browser or AT and after > doing a lot of testing with XP, Jaws, IE and FF the available list > was very very short. I believe one of the ideas going into the DHTML > style guide was that ARIA would allow an AT to know that the user > had focus on a widget and get out of the way. Without that it would > be nearly impossible to find a set that works cross-AT, cross- > browser and cross-platform. Do Mac Voiceover users care that control- > J jumps cells with Jaws on Windows? Do Windows users care that > voiceover users jump between headers using control+alt+h? I suggest > that the set of available key combinations that are as agnostic as > the web sites we want to implement them on is nearly null. In light > of that, a clean slate approach seems appropriate. Given no > constraints on keystrokes other than trying to give a nod to what is > common (familiar) in existing implementations to lower cognitive > load, what would make the most sense for navigating and controlling > widgets? For what it's worth, I completely agree with you. The argument I've heard against that is that there needs to be a consistent mechanism for keyboard navigation even if not controlled by AT like a screen reader. To that, I replied that the user agent should implement the key commands. For example, I can activate the menus or form controls in Safari with or without VoiceOver. The same should be true of all ARIA widgets in that UA and AT control web application widgets in the exact same way as the desktop equivalents. Although I firmly believe this is the right approach, not all browsers currently support DOM mutation events properly, and that feature is required for this approach to be a practical solution. At this point, I've started mostly ignoring the DHTML Style Guide as an overly-complex, but nice-to-have stop-gap measure until user agents support all these controls natively. I'm not saying this to make enemies in the DHTML Style Guide Working Group, but that will probably happen anyway. I'd be more on board with the more simple approach Victor mentioned: very basic navigational controls, including the keystroke to open a contextual menu that contains all the more-complex methods of navigation and control. James
Received on Wednesday, 19 November 2008 20:23:38 UTC