- From: Charles McCathieNevile <charles@w3.org>
- Date: Tue, 24 Aug 1999 22:47:13 -0400 (EDT)
- To: Al Gilman <asgilman@iamdigex.net>
- cc: w3c-wai-ua@w3.org, w3c-wai-gl@w3.org
As a refinement to general strategies for naviagtion I like it, although I am loath to recomend it as something that should be done instead of correcet use of markup, since the effective message is that it is pointless bothering with structured markup since effective strategies don't use it. Thinking some more: Title is human readable metadata about almost any element. As such it should be renderable (at the discretion of the user. So we can either use it to bypass the role of strongly typed data, such as HTML marked up accroding to the intended meanings of the elements as laid out in the specification, by using the title as a means of directly communicating verbally the same semantics which we use visually. This seems to me an equivalent to reverting to HTML 3.2 presentational markup, on the basis that we can now provide a non-visual equivalent. For HTML this seems a bad strategy. (The rules change a bit in XML were you can declare your own structure, and you declare separately how the various elements ought to be rendered. This allows you to declare something based on your own understanding of visual semantics, and structure it according to the logic behind those semantics. The structure is made explicit in the structuring of the XML DTD/schema/profile. I suspect that in writing presentation based, poorly structured HTML, people are selecting the particular elements they want from a given set. So starting from an empty set, and creating types of things could induce people to provide more structured semantics in a more natural way.) So I am still a fan of map, which is an HTML element created for the purpose of holding a collection of links. Charles McCN On Tue, 24 Aug 1999, Al Gilman wrote: At 04:43 PM 8/24/99 -0700, Jon Gunderson wrote: > >Option 1: >Using the DIV element and some type of CLASS or NAME identifier: >PROs: Easy to implement and author >CONs: There is no mechanism to reserve class names in HTML, there could be >conflicts if specific class names are used Variation: the TITLE and DIV approach. After today's coordination group call, Len came up with the idea that perhaps the TITLE attribute is a key resource. If authors do the right thing with the TITLE, maybe we don't need the UA to recognize some particular CLASS to do the right thing. Let's see what we can do if we pursue that idea. The basic dodge is that for things you want to move high up in the navigation structure, you don't TITLE too many of their ancestors in the parse tree; and for things you want to hide or minimize you add DIV structures with TITLEs so that the collection gets listed rather than the members (unless the user explicitly enters the collection). The User Agent method that goes with this markup strategy is hierarchical navigation of an Effective Table of Contents tree. The Effective Table of Contents tree is the parse tree except that containers that have no TITLE attribute on them are ignored or equivalently flattened. Elements below them act (in the effective tree) as though they were directly descended from the next TITLE-bearing ancestor up the tree. The trick here is that DIV, on the other hand, lets us push things _down_ the tree. We can encapsulate and quickly skip the page masthead, while the TITLE technique lets us ignore non-semantic superstructure like layout tables. The attached three files are 1) the original web page from Amazon.com 2) the modified page with TITLE and DIV added according to this technique and 3) the top level of the resulting effective table of contents. Think about it. Al --Charles McCathieNevile mailto:charles@w3.org phone: +1 617 258 0992 http://www.w3.org/People/Charles W3C Web Accessibility Initiative http://www.w3.org/WAI MIT/LCS - 545 Technology sq., Cambridge MA, 02139, USA
Received on Tuesday, 24 August 1999 22:47:14 UTC