W3C home > Mailing lists > Public > w3c-wai-gl@w3.org > July to September 2004

RE: Linear reading order

From: Jason White <jasonw@ariel.ucs.unimelb.edu.au>
Date: Wed, 7 Jul 2004 21:42:09 +1000
Message-ID: <16619.57745.229238.577862@jdc.local>
To: "Yvette P. Hoitink" <y.p.hoitink@heritas.nl>
Cc: <w3c-wai-gl@w3.org>

Yvette P. Hoitink writes:
 > 
 > However, we could require that _if_ reading order is predefined, it should
 > be logical. 
 > 
 > Using HTML+CSS, I can do stuff like
 > 
 > <p style="float:right;">no sense.</p>
 > <p style="float:left;">This sentence makes </p>
 > 
 > which will make it visually look like "This sentence makes no sense" but
 > will be read by screenreaders as "no sense. This sentence makes" (Go Yoda!).
 > 
 > 
 > I think cases like these are bad for accessibility and we should have a
 > success criteria somewhere in our guidelines that addresses this issue. I'm
 > thinking along the lines of:
 > "If a reading order is provided, make sure it is logical". 

Let me first agree that the above is an excellent example of what we
don't want, and that it is bad for accessibility by anyone who isn't
using the default CSS media type for the document (e.g., "screen").

Now regarding Yvette's proposal, it very much depends on what is meant
by "providing" a reading order. In the above example, a reading order
is, in fact, provided, albeit implicitly in the presentation created
by the CSS in the STYLE attributes. Thus it would be necessary to
define what "providing" a reading order amounts to.

Perhaps what we want is rather that the order and hierarchy of the
markup should reflect the logical structure so that tree traversal
operations will yield one or more logical reading orders. The
paragraphs in the above example are siblings in the tree structure,
one to the right of the other, and what we want is that the
left-to-right traversal of the structure gives a reasonable reading
order in this case.

Another example which has been cited before is that of an SVG image in
which the markup is optimized for rendering so that the whole
structure of the image is lost. Query whether this is the sort of
practice that should be discouraged where metadata specifying the
logical structure is not also provided. I think the example was of a
coloured diagram in which all the lines in one colour were
represented first, followed by all those in the next colour, etc., to
optimize the markup for the rendering algorithm.

I don't have any solutions to offer at present.
Received on Wednesday, 7 July 2004 07:42:21 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 5 February 2014 07:17:58 UTC