Re: TABLE guidelines

> 7.2 Avoid using tables to arrange text documents in columns.

This item needs to be reclassified as an INTERIM measure. 
Hopefully screen readers in the near future will be able 
to look at the browser's document object model and read 
down a column, cell by cell, etc. through a table.

In addition, the point needs further clarification.
For example, using the TH element to produce table
headers is a good idea. However, text headers
will also produce multicolumn-screenreading problems.
Using the hack of putting BRs at the end of each
table cell changes the accessibility of multicolumn text
for Lynx but not for Navigator. Does this need
to be added as an interim guideline as well? 

What 7.2 is really saying is that for specific
combinations of screen-reading software and browsers,
multicolumn text of any kind is a problem. 
In order to give 7.2 the precision it deserves these problem
combinations of browser and screen reader must be 
specified in the guidelines albeit maybe the longer document.

Further, it would seem that one additional step would be 
to generate statistics on the relevant combinations of 
browser/screen-reader software so that we can judge 
over the next 2-3 years when to phase this and other 
interim points out.

Josh Krieger
CAST

Daniel Dardailler wrote:
> 
> I was rereading WD-WAI-PAGEAUTH-0203 in light of some real life
> application of the guidelines and I have a question regarding the
> requirements on table.
> 
> What's the difference between:
> 
> 7.2 Avoid using tables to arrange text documents in columns.
> and
> 7.4 Avoid using tables to layout a page.
> 
> one is required, the other not, they both refer to layout of documents
> using table construct. We should be more detail if possible.
> 
> 7.1 is a REQUIRED about
>  "Associate table cells with row and column labels explicitly"
> 
> This is about adding 'header' or 'scope' to the TABLE markup.
> 
> I wonder if a default algorithm (or set of default values for these
> attributes, if you prefer) could not be described that would alleviate
> the needs for updating all the table markup out there.
> 
> For instance, the TABLE example number 2 in the appendix adds
> scope=col to a bunch of TH but it looks to me that a user agent can
> very well infer that data from the fact that it's a bunch of TH in a
> TR followed by a bunch of TD in TRs.
> 
> I kind of remember Dave Raggett describing such an algorithm for user
> agent to follow. I need to look back in the archive.
> 
> I wouldn't want to make anything that can be achieved by a user agent
> a REQUIRED.

Received on Wednesday, 4 March 1998 17:46:31 UTC