Robert J Burns wrote:
Whatever the HTML5 draft says for document conformance, the use of the headers attribute to point to a TD cell is long established and much existing content will be broken if the HTML5 headers association algorithm does not support it. The question of whether document conformance should permit a headers attribute to reference an TD is a separate matter. Personally, I'm finding it difficult to come up with an example where the approach you advocate wouldn't work. The question is how important is it for authors to distinguish not only data from header, but also data acting as header (or header acting as data to put it another way).
Also we would need to work out how UAs behave when they encounter a proper <th> (or several) embedded or nested within a complex data table. Would it confuse existing AT? Will the UA think that it is encountering a separate table and start a new heuristic process? How will the presence of nested tables therefore effect the way the UA reacts?
I haven't tested any nested headings within tables but James may have. I would be concerned that using nested headers in data tables this way could be more problematic that the use of conceptual headers.
My 2 cents (Euro). Cheers Josh
