Re: Why I don't attend the weekly teleconference (Was: Input on the agenda)

Steven Faulkner wrote:
> not that I know of, commercial vendors AT are the opposite of open.
> They consider what's under the hood of their products to be commercial
> secrets. To understand much of how AT behave and what they support,
> reverse engineering is required. Which is what myself and others have
> done for many behaviours.
> I haven't done so for their table filtering as I have other testing
> prioties (ie ARIA) and also don't consider that it will make a
> difference in this particular debate whether the algorithm is
> understood or not.

It seems to me that knowing about the efficiency of layout table 
filtering is critical to arguments of the form "harmful summaries on 
layout tables are not problematic because UAs filter them out". This 
does not imply that the exact algorithm must be known, but it does imply 
that the various probabilities like P(table is identified as a layout 
table | table is a layout table), P(table is identified as a data table 
| table is a layout table), and so on, are necessary to know what

As it happens, Mozilla have an open source layout table detection 
algorithm [1] which is rather simple. Sadly it needs computed style (and 
hence a browser) to fully implement. This makes it harder to run in an 
automated way over a large set of tables.

[1] 
http://mxr.mozilla.org/mozilla/source/accessible/src/html/nsHTMLTableAccessible.cpp#1033

Received on Tuesday, 30 June 2009 12:05:12 UTC