RE: mobileOK Pro 1st Draft

Le vendredi 07 mars 2008 à 14:06 +0100, Scheppe, Kai-Dietrich a écrit :
> > * Access keys: 
> >         - "Where there are elements, particularly navigation links and
> >         form controls, that would benefit from access keys:"
> > How does one determine that there are such elements?
> 
> By looking at the content

OK, let me rephrase my question, then; if tester A thinks a given
element needs an accesskey, and tester B thinks it doesn't, how do I
determine who is right?

> >         - "if access keys are not indicated effectively":
> > What does it mean to be indicated "effectively"? For 
> > instance, is it enough to have a page on the site that lists 
> > you access keys? or do they need to be indicated on the page itself? 
> 
> That depends on the context. In some cases link decoration may be
> needed, in others a page where usage is explained or some other
> mechanism.
> Important is, that the user knows which access keys can be used or find
> out about it.

I fully understand the principle behind the test; I'm worried here again
about how this test can be applied in practice with a sufficient level
of uniformity.

> >         - "If the usage of access keys is not consistent 
> > across a given
> >         page and site"
> > How do you determine what constitutes a site? How many number 
> > of pages in the site to you need to check to determine this 
> > consistency?
> 
> Since mobileOK is based on a assertion made by a content provider or
> author the scope of that assertion has be to be made clear.
> POWDER is a vehicle by which this can be accomplished.

OK, that makes sense; this should probably appear somewhere in the
document.

> Consistent means without change, so everywhere within the scope.

Well, "without change" is not what the BP implies, at least.

> [...]

> Precicely, which is why a human being has to make this test.
> [...]
> The question is how far do we want to minimize variation?

I think there are three degrees of testability:
 * machine-testable
 * objective human-testable
 * subjective human-testable

I'm worried that mobileOK Pro is using a lot of subjective human-tests,
which means that the notion of being mobileOK Pro is going to be
extremely subjective, and thus of a fairly limited value.

(not to mention that we've been warned by the W3C Advisory Board to be
very careful with subjective testing:
http://lists.w3.org/Archives/Member/member-bpwg/2006Aug/0182.html)

Dom

Received on Friday, 7 March 2008 13:22:48 UTC