In GMail and Calendar, Google seems to be providing accessibility support through two models, neither of which is fully functional or implemented. First, they are providing shortcut keys for almost every function. Second, they provide limited semantic markup in the form of ARIA landmarks, headings, and other functional blocks to aid in page navigation and usability. The shortcut key model provides a lot of helpful navigational and functional aids to many assistive technology users by allowing these users to avoid excessive “tabbing” to navigate and use the application. With screen readers though, these two models bring into focus some of the fundamental issues at hand through three main problems.
First, many screen reader users will want the ability to interact with the application in standard ways utilizing the semantic structure of the page, and that level of support is not present in either GMail or Calendar. Second, the keyboard shortcut model is not reliably communicating results of actions back to the screen reader. The only screen reader that voices any confirmation of a keyboard action is ChromeVox. It is unknown at this time how the applications are communicating that information back to the screen reader, whether it is through proprietary or non-standard methods only available to Chrome and ChromeVox or through ARIA attributes that other screen readers do not support yet. The combination of JAWS and Firefox provides a lot of support for ARIA attributes so this case seems unlikely. Third, many of the shortcut keys interfere with the operation of the screen reader because there are already screen reader specific keyboard shortcuts assigned to those key combinations, thus requiring the user to perform screen reader acrobatics in order to effectively use them. (It should be noted that there are ways to remap shortcut keys to different key combinations that will not conflict with the screen reader, but since there are so many keyboard shortcuts this is an impractical solution for users.)
While there will be continued debate over which accessibility model is most appropriate for delivering accessible Web applications, it cannot be denied that both models lack sufficient support to allow screen reader users to effectively use the applications. As Google adds more accessibility support, we encourage them to fully implement support for both accessibility models so that all functions can be accomplished through either model, exclusive of the other model.