Representing document reading order in HTML

I fully concur with Jon Gunderson's comment with respect to reading order:
some standard conventions need to be defined, taking advantage of the HTML
CLASS and ID attributes where necessary, which will enable the author to
specify the linear reading order of the document. To some extent, this
requirement is already accommodated by HTML 4.0, for instance in the
TABINDEX attribute which defines the sequence in which the user is to be
moved through the entry fields of a form. Although the HEADERS, SCOPE and
AXIS attributes used in tables do not directly define a reading order as
such, they make explicit the semantic connections between header and data
elements, thereby supplying the infrastructure on which style sheets or
client-side scripts can build so as to read the table in a comprehensible
fashion.

The need to vary the reading order of a document has often been stated as
a requirement, but remarkably few examples have been given. What is needed
at this stage is a taxonomy of examples out of which a solution can be
developed. The guidelines could then suggest appropriate CLASS values
which could be associated with HTML elements, especially the generic
containers (SPAN and DIV) to provide the user agent with the needed
structural information. If a more generalised approach be desired, then
presumably the suitability of RDF mechanisms for this purpose could be
explored by the PF working group.

The primary example which comes to mind is that of a navigation bar. It
would be desirable to move such a construct to the end of the document, or
enable the braille or audio output software to notify the user of its
presence without reading its contents in detail, unless the user chooses
to do so. The guidelines already give an example of appropriate markup in
this case. Perhaps a better CLASS value would be: CLASS="navigation"

Presumably, tables and long descriptions (for instance the contents of
OBJECT elements) constitute other components of the document which the
user may want to defer during the reading process, but these are already
indicated by their own markup and thus the issue becomes solely one of
user agent design. Perhaps other members of this working group could
furnish additional examples at this point which would help to clarify the
scope of the requirement.

Received on Wednesday, 6 May 1998 22:12:39 UTC