Re: operational concept for table browsing

to follow up on what Jason White said:

> Having considered the various ideas which have been set forth during the
> past few weeks, it seems clear that little advance has been made upon the
> current HTML 4.0 draft. As I think all participants agree, AXIS and AXES
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> should remain the basis of table accessibility. Al has indicated that it
> may be necessary to offer a mechanism for encoding the semantic
> relationships among data cells, which extends beyond the capabilities of
> the AXES attribute as currently defined. However, neither Al nor any other
> WG participant has yet produced a convincing example of how such coding
> would improve the audio rendering of a table, and thus the case for it has
> not been convincingly made out.

I had a good conversation with Dave Raggett today.  I expect yet 
tonight (next three hours) to write up some more of this.  We have 
neither reached consensus (between even just the two of us) nor
impasse.  We are closer as a result of today's discussion.

I also have a chance this coming weekend to debrief a browser
author who has done some work with table-browsing functions and I
expect to be wiser about what matters and what doesn't after
this.

So please bear with me.  We are making progress and it is not time
to rush to judgement on this one.  This is the toughie in the whole
subject area.  I will try to delineate what I think is agreed and
what is open.

One of the things that Dave and I agreed was that we should be
looking at examples.  Please nominate tables to be used as case
studies.  If the table has some way such as a list in a textonly
companion page, that is good.  But we want some inaccessible
examples to see if we can resurrect them, too.

> It has also been suggested that CLASS values be reserved by which the
> author can indicate to a user agent whether a table is to be read, by
> default, across the rows or down the columns.

Yes, further discussion is in order on what classes should be
commonly recognized.  In addition to the above two, the standard
data modeling SET, BAG and LIST are candidates.  One might want
to call the above two COLSET and ROWLIST.  There is an
intermediate class that could be called OTHER where there are
some links between cells but they do not follow a pattern which
is consistent across the entire depth or breadth of the table.

-- Al

Received on Thursday, 9 October 1997 21:28:32 UTC