On 30/1/09 15:14, Willie Walker wrote:
Determining if there are landmarks can be an expensive operation and can slow performance down. So, it might be best to only do landmark operations when the user explicitly requests them. In any case, what I'm hearing is that you are joining the list of users who want default keybindings. Do you think the 'm' key is reasonable for this?
I realise JAWS is providing a dedicated interface for landmarks, but it seems ultimately wrong to treat landmarks differently to other sources of semantics, whether that takes the form of explicit markup or is derived through other heuristics.
End-users should be able to deal with document and application concepts, rather than having to worry about the underlying markup.
For example, WAI-ARIA includes the "navigation" role, HTML5 includes the "nav" element, and in their absence a user agent could make a reasonably good guess about what constitutes the navigation on a page with pattern matching. Or again, WAI-ARIA includes the "main" role, HTML5 includes the "article" element, and in their absence a user agent could make a reasonably good guess about what constitutes the main content of a page. I don't think those three sources of the same information should be presented to the user any differently, but especially not the explicit DOM sources. Why should the role be exposed to the end-user with a different user interface than the element? If users wanted to know about the DOM structure, they'd use a DOM inspector.
There does seem to be something broken somewhere in the architecture if querying for landmark roles is such a heavy performance cost; it's not a heavy cost with XPath on the DOM.
-- Benjamin Hawkes-Lewis