W3C home > Mailing lists > Public > public-html@w3.org > August 2007

Re: Heuristic Tests for Data Tables (Discussion)

From: Ben 'Cerbera' Millard <cerbera@projectcerbera.com>
Date: Thu, 30 Aug 2007 03:22:30 +0100
Message-ID: <004501c7eaac$9f9e1380$0201a8c0@ben9xr3up2lv7v>
To: "Philip Taylor" <philip@zaynar.demon.co.uk>
Cc: "HTMLWG" <public-html@w3.org>

Philip Taylor wrote:
> Are there any details on how existing implementations distinguish between 
> layout and data tables?

Some quick searches of the web turned up a lot of documentation for 
accessibility testing software. For example:

* <http://www.contentquality.com/About%20Reports/DataTables.htm>

Responding to my own speculation about identifying layout tables in 2005, a 
representative from the "Silktide" website scoring tool indicated they apply 
a range of tests which are added up and compared with a threshold amount:

* <http://www.accessifyforum.com/viewtopic.php?p=31429#31429>

Reportedly, <th> causes confusion for "screen reader users" (that's too 
vague for me to check up on) when in a layout table:

* <http://www.webaim.org/techniques/tables/#header_tags>

I found some specification documentation about JAWS from the vendor's 
website, but it's for a very old version (4.52):

* <http://www.freedomscientific.com/fs_support/BulletinView.cfm?QC=406>

JAWS can now be told how to determine which tables on the web should be 
treated as data tables via the default.jcf file. This allows you to 
customize four variables:

a) the minimum number of cells that a row must have in order to be 
considered a data row

b) the minimum number of data rows that a table must contain

c) the minimum number of rows in the entire table which must contain some 
text and

d) the minimum number of columns in the table which must contain some text

I imagine the popular screen readers are now more sophisticated than merely 
treating tables with few cells as layout tables. Then again, maybe that's 
all which is needed in practise? I'm not having much luck finding 
authoritative sources about this.

Philip Taylor wrote:
> I guess a first approximation would be to see whether the table contains 
> any block-level or image elements, in which case it is more likely layout, 
> but I don't know if more subtlety is needed.

Legitimate data tables may include images [1] and layout tables may contain 
no images [2]. Block-level content can appear in data tables [3] and might 
not appear in layout tabes [4].

A 100% accurate algorithm may be impossible but a bit more subtlety would be 
worth looking into. If AT vendors get more involved we might find this is 
already a solved problem.

[1] <http://sitesurgeon.co.uk/tables/clark2006/06-gui/original.html>
[2] <http://www.sitemorse.com/clients.html>
[3] <http://www.w3.org/QA/TheMatrix>
[4] <http://azz.gouranga.com/>

Ben 'Cerbera' Millard
Collections of Interesting Data Tables
Received on Thursday, 30 August 2007 02:22:50 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 29 October 2015 10:15:25 UTC