Re: What is a subject of a test?

On Fri, 01 Apr 2005 23:35:35 +1000, Johannes Koch  
<> wrote:

>  For me an assertion is about whether a subject passes a test. If I say,  
> the subject is the whole resource, than there should only be one  
> assertion containing multiple locations. If OTOH I say, the subject is  
> e.g. the markup element, then there should be two assertions containing  
> one location each.

If we go down this path then we either need to explain how to deal with  
the cases I mentioned that make it difficult:

1) A page has two img elements which lack alt attributes. This is two  
failures of the same test, on the same page.

2) A page has an image map, and elsewhere, has redundant text links for  
the image map. Both of theselocations need to be specified.

The second case already suggests that it is hard to go with the model of  
the test subject being a part of the page, not the page itself.

Or, we can follow the approach of making the subject a part of a page. But  
how do you then define what the page as a whole is, and whether it passes  
WCAG, based on a whole colection of tests of part of it?

This can now be done in standard RDF, but it is pretty painful, and I  
don't see the value of it. It means that every item in a page needs to be  
described seperately. And that also runs into the problem mentioned in the  
second of my dificult cases, where for some tests you describe two things  
at once, and for others you describe a single thing.

In essence, two problems noted means two assertions. The point of  
seperating out the locator from the subject explicitly is that it becomes  
trivial to look at an Assertion and see if it has the same subject as  
another assertion, without having to process a list of the component parts  
of a subject in order to determine whether two subjects of Assertions are  
parts of the same test subject or not. At the same time, it is easy to use  
the locator information.

As an added bonus, this design allows you to have different types of  
locator information according to the test subject type, and/or the  
testcase, and/or some other factor. Which is going to help with the  
persistence problem...



Charles McCathieNevile                      Fundacion Sidar   +61 409 134 136

Received on Friday, 1 April 2005 14:33:06 UTC