- From: Anne van Kesteren <annevk@opera.com>
- Date: Fri, 21 Sep 2007 16:25:38 +0200
- To: "Aaron M Leventhal" <aleventh@us.ibm.com>, public-html@w3.org, wai-xtech@w3.org
On Thu, 20 Sep 2007 17:24:55 +0200, Aaron M Leventhal <aleventh@us.ibm.com> wrote: > Please take a look at the FAQ. I hope it will help frame a constructive > discussion about ARIA. One particular part I want to discuss in this > group is Ian Hickson's proposal (in the FAQ) to remove the namespace > dependency > for ARIA properties. I have run this proposal by other browser > manufacturers and large content providers involved in ARIA development. This seems like the best way forward, indeed. Here's a bit more detailed on how I think it could work taking into account legacy content, ease-of-authoring, consistency with other standards, and deployed software: role attributes on elements in the http://www.w3.org/1999/xhtml namespace are supported and take a space-separated list of roles. (This covers both text/html and application/xml for what it's worth if the parsing algorithm in HTML5 is followed.) role attributes in the http://www.w3.org/1999/xhtml namespace on elements not in the http://www.w3.org/1999/xhtml namespace are supported and also take a space-separated list of roles. ARIA properties are supported as aria-{property} attributes on elements in the http://www.w3.org/1999/xhtml namespace (replacing {property} with the name of the property). ARIA properties are also supported as {property} in the http://www.w3.org/2005/07/aaa namespace on _all_ elements. These take precedence over the aria-{property} attributes on elements in the http://www.w3.org/1999/xhtml namespace. (Equivalent to how the lang and xml:lang attributes interact.) Note that it is impossible to write these down in a text/html serialization unless it is done in the DOM that is generated from such a serialization. As for the list of roles, I would prefer if we simply remove the complicated qname idea and make them opaque strings just like the rel, class, etc. attributes already work. This seems more consistent and easier to author. Especially considering that the DOM and CSS don't have support for namespaced attribute values so that authoring once, deploying everywhere (for equivalent documents) does not work automatically. This seems problematic. -- Anne van Kesteren <http://annevankesteren.nl/> <http://www.opera.com/>
Received on Friday, 21 September 2007 14:26:11 UTC