W3C home > Mailing lists > Public > public-wai-ert@w3.org > March 2005

Re: EARL and describing tests Re: Agenda for F2F at TP

From: Charles McCathieNevile <charles@sidar.org>
Date: Tue, 08 Mar 2005 10:01:58 -0500
To: jim@jibbering.com, "Karl Dubost" <karl@w3.org>
Cc: public-wai-ert@w3.org
Message-ID: <op.snbqpkf2w5l938@researchsft.witness.org>

I think Jim is right that we need to be able to identify tests, and  
provide at least basic descriptions of them.

We also need to describe how a set of tests relates to a different set of  
tests in terms of the fact that two different tests or sets might imply  
something about the result of a third test. This is not about testing  
process, but about the equivalence between results - for example passing  
Hera's test for text alternatives on images implies that you will pass  
Chris' test that there is an alt attribute for each img element.

I think both of these can be done by virtue of the fact that EARL is RDF,  
using the basic framework stuff it gives us (RDF Vocabulary Definitions,  
OWL, etc) rather than having to define anything in EARL itself. This  
strikes me as a good thing...

cheers

Chaals

(Lots of context below)

On Tue, 08 Mar 2005 07:26:27 -0500, Jim Ley <jibberjim@gmail.com> wrote:

> On Mon, 7 Mar 2005 16:20:57 -0500, Karl Dubost <karl@w3.org> wrote:
>>
>> Le 28 févr. 2005, à 10:43, Charles McCathieNevile a écrit :
>> > Hmm. EARL provides some simple (and useful) ways of describing tests.
>> > An example is the descriptions that can be extracted from among other
>> > information in
>> > http://www.w3.org/2001/sw/Europe/200305/axforms/earlinst.rdf - the
>> > relevant parts are as follows:
>>
>> I do not recommend to use EARL for describing test. I would focus
>> rather on an input interface to reference the test.
>>
>> EARL should really be reserved for the reporting. A Test suite is not
>> only a collection of Test Cases, but test cases, manual, a framework to
>> pass the test, etc.
>>
>> I think using EARL to describe the test would be a mistake, because it
>> will make it suddenly a monster.
>
> Almost certainly true, but without a mechanism to describe tests even
> at the most abstract level, I can't see how EARL can be particularly
> useful, someone has to define the tests, and whilst I don't think we
> should do more than provide a framework on how to do it (one IFP and a
> suggestion of a NAME/DESCRIPTION property to use) I do think that is
> essential.  Otherwise I can't see how we can consume EARL from other
> applications.

-- 
Charles McCathieNevile                      Fundacion Sidar
charles@sdar.org    +61 409 134 136    http://www.sidar.org
Received on Tuesday, 8 March 2005 15:04:13 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:18:25 GMT