The Future of AccesskeysPosted: August 14, 2004
The July 22, 2004 Working Draft of XHTML 2.0 abandons accesskey for a much more powerful and flexible access attribute that will put the user in control and allow keyboard users to define their own keystrokes.
For over a year we've written at length about some of the problems with accesskeys. It started with our piece Using Accesskeys: Is it worth it? and has continued from there. We've done a lot of research on accesskey availability, found keystroke conflicts, and proposed alternatives to accesskeys that put the user in control, rather than the author.
During that year, I can't count the number of times that I've written statements like this at AccessifyForum.com, on mailing lists and in comments on blogs - one of which was on Dave Shea's post "I do not use accesskeys."
- We like the idea behind accesskeys,
- We find the implementation very problematic
- We believe that we collectively need to examine the issue and determine if we need to develop other solutions that better suit the needs of everyone.
Through all of those articles it boils down to a few key points:
- authors (usually) define (too many) accesskeys
- accesskeys are used inconsistenly across sites, meaning they are of generally limited use
- accesskeys cause conflict with other keystrokes
- current user agents do not allow accesskeys to be turned off, nor defined by users
We often re-examine our thoughts on accesskeys. We believe they can be of some value -- most likely used in well-defined web-based applications like WordPress or Basecamp. However, even when they are in use in these environments I still don't have the option of turning them off, nor can I override them with my own keys. I believe that there just isn't enough flexibility in current browsers to make accesskeys as useful as they should (or could) be.
A new strategy
It appears that a new strategy has emerged. Instead of defining "keystrokes" with the accesskey attribute, the XHTML 2.0 Working Draft (July 22, 2004) takes a big step forward to addressing the issues with accesskeys by dropping the accesskey attribute and introducing the access attribute. This is new in the July 22, 2004 Working Draft. The previous version (XHTML 2.0 Working Draft (May 6, 2004)) did not have the access attribute, and still made reference to the accesskey attribute.
This will allow authors to define access points instead of keystrokes.
Here's what we see as the most important point from the working draft:
The invocation of access shortcuts, and the assignment of key or other bindings to them is not further defined here. A user agent should allow the user to access the user agent's list of recognized access points, to add to them, and to specify bindings for them.
This significant change means a few things:
- Authors will no longer have to worry about keystroke conflicts because it is no longer their job to define keystrokes. Authors will also need to adjust to the fact that they no longer have control.
- Authors can focus on defining logical and meaningful access points in their documents
- Users that really need and could benefit from this functionality can define their own keystrokes, eliminating/minimizing conflict with their software.
- Users get consistency across sites because the keystrokes are defined at the browser level. No more remembering what keystrokes do what on which sites.
Anyone who has read our other articles on accesskeys will hopefully recognize that we are incredibly happy with these changes - they closely resemble what we've been hoping for and writing about for the past year, albeit in a slightly different format.
Recognized access points
This strategy requires clearly defined access points. Perhaps the well-defined link relationships and expanding on those as needed would be a good place to start. These would include obvious access points such as Search, Home, Help, Contact, and could be expanded to include others such as content, form, accessibility, privacy, navigation and others. We discussed this earlier in Link Relationships as an Alternative to Accesskeys
The working draft notes:
A list of access points (such as 'main content', 'navigation list', 'submit button' is in preparation.
My instinct tells me that single word access points would be better, though I can't say for certain that one word would always be appropriate. "Submit button" could just as easily be "Submit". (One other note regarding the use of "Submit" as an access point - I'm assuming that the behaviour would be to focus on the submit button rather than actually submit the form, otherwise we'll see obvious problems on pages with more than one form.)
This working draft recognizes that keyboard users are best equipped to edit and add to the keystrokes they need is a very big step forward, and fits with most of what good design and standards is about. People, and putting people in control of their own environment. I'm sure my friend Patrick Lauke (redux) will be as happy as we are with this one -- the onus is on both the author (to define the access points) and the user (to do something meaningful with them).
This strategy allows me as the user to define keytrokes that work across all sites that implement these access points, and it allows me to create a highly customized set of keystrokes for working with various web applications that define additional access points. If I want Ctrl + Shift + H to take me to the home page, and Ctrl + Shift + S to take me to the search form or page for a site, then it will work on any site where these access points have been defined.
Access > Accesskey
Looking forward, it might even be useful to allow for user agents to allow loading of custom keystroke profiles so that users can create and save different sets of keystrokes for use in different applications. That way I could easily define keystroke mappings for WordPress and another set for Basecamp. Now that would be useful. Site authors could also provide an XML version of the access points defined with the access attribute so that users could import the access points directly into one of those custom profiles.
We believe that this new strategy provides much more power and is much more useful than the current accesskey. Unfortunately it may be some time before we realize the benefits of this change in strategy, as XHTML 2.0 may be a long way from the mainstream, and in order to make this strategy successful, user agents need to change to allow this level of customization. The Opera browser is already a leader in this respect as they allow the user to create highly customized keystrokes and mutliple keyboard configurations through their Preferences interface. Let's hope that other user agents can catch up.