[html4all] Better AT user testing of table features

[Leif Halvard Silli]

Hi Advocates,

The table tests that Joshue published are, as I understood it, compare the user experince of different tables with different «semantic styling». But to really compare the user experience, one should test the very *same* table(s) with different «semantic styling». Another critisism of the tests is that because of what I said above, they don't test the difference between the algorithms of HTML4 and (current) HTML5 - and thus we can say little *general* about either or with regard AT UAs. Therefore I propose a real, qualitative test.

James Graham has made this great [table inspector] with which one can test interactively the headers cell relationship. The name of all the header celss a cell has, is listed inside each cell. And in addition, by clicking the cell, its headers cells are highlighted in another color. The table inspector lets you test the same table in in the semantic styling (= the algorithms) of HTML4, HTML5 and 2 other, experimental, algorithms.

I think it would be great if one could perform the same test in JAWS and similar. For such testing, the relationships of each algorithm should be hardcoded via the headers="" attribute. Because, the idea is to test the user experiences - and not to test how good AT software we have.

We could either ask James Graham [which I just did] to update his table tool so that it could (optionally) hardcode the relationships with headers="" references. That would be the best. But in any case, or in addition to him updating the tool, it could interesting to provide a test results from a user testing based upon the same table/tables read with different algorithms.  

What do you think? Especially what do our «test masters» Steve and Joshue think? And should I make the proposal on the public-html list? (Then e.g. James Graham would understand my request to update his tool better, I suppos.)

The table test tool really reveal difference between the HTML4 algorithms and the other algorithms. Consider a "normal" table with 4 TH-cells in the first row: TH1, TH2, TH3, TH4. In HTML4, TH1 is header cell for TH2, while both TH1 and TH2 are headers for TH3. And TH1, TH2, TH3 ar headers for TH4. 

But are those differences useful for AT users?  For testing, I have used the [ Complex Table examply by Rob ]. And I think, when we look at the amount of repeated text in each cell, which the HTML4 algorithm creates, it can look overwhelming. But this might be exactly what the unsighted user need. And this would therefore be interesting to document.

[Anne's table] was mentioned in one of the message from James. That table uses @SCOPE and seems to work worse in the HTML4 algorithm than in the other algorithms which the tool offer.

[table inspector] <http://james.html5.org/tables/table_inspector.html>
[ Complex Table examply by Rob ]
<http://lists.w3.org/Archives/Public/public-html/2007Aug/att-0484/ComplexTables.html>
[ Anne's table ] <http://annevankesteren.nl/2007/09/tmb-overview>
-- 
leif halvard sili