Re: TABLES - To label or not to label

> Question 1: Should a guideline state that labeling of simple tables is not
> [required]?

yes 

> If yes, then:
> 
> Question 2: When does a table become complicated enough for labeling to be
> [required]?

This is of course the tough question.

I talked to Dave Raggett about this issue and I think we both agree
that there is a threshold below which a table can be considered
"simple".


I propose the following:

   - if it only has a single row of header cells at the top
    and/or
   - if it only has a single column of header cells on the left side
     (modulo right-to-left directionality)

In these two cases (which should be treated together), given a cell,
it is really easy for an agent to find the column and the row header
they belong to without requiring the author to add more markup.

In other words, there is a potential "default" value for the scope
attribute on TH that can be defined as: 
   - if all the cells in a row (TR) are header (TH), 
         then scope defaults to col 
   - if only the first cell in a row (TR) is a header (TH)
         then scope defaults to row

We might go as far as trying to define default values for rowgroup and
colgroup cases (covering sub-headers) but I'd be happy with just that
in the guidelines.

> Note 2: labeling refers to the new TABLE markup available in HTML 4.0 such
> as header, id, scope, and summary

summary and abbrev should be RECOMMENDED regardless of the above, they
do no serve navigation but description.

Received on Wednesday, 4 March 1998 08:36:30 UTC