- From: Shadi Abou-Zahra <shadi@w3.org>
- Date: Wed, 30 Mar 2005 22:38:19 +0200
- To: "'Johannes Koch'" <johannes.koch@fit.fraunhofer.de>, <public-wai-ert@w3.org>
Hi Johannes, I'm thinking out loud: Missing alt texts could be addressed by something like "//img!@alt" (not sure if that is correct XPath). In such a case, it is probably better to have one assertion per test case (because such a generic pointer supports some sense of persistency). However, if the error is a "bad" alt text, then we would probably need to regard the value of the alt attribute too (e.g. "//img@alt='stupid text'"). There are occasions when such an expression could match more than one occurrence, (for example if the evaluation tool can be configured to alert when it encounters the string 'stupid text' as an alt text) but in end effect we may be generating more than one assertion for a single test case. Finally, if we can not use such XPath-type expressions (for example if the page is not well formed [enough]), then the Subject probably has the be the URL of the page (i.e. we can not differentiate the location for the different instances). So it seems that ideally the Subject is [something like] an XPath if the document is well formed, or a simple URL as a [compatible] fallback. The number of assertions which are generated depends on the test itself (because that could also effect confidence, error message, etc). What do you think? Regards, Shadi -----Original Message----- From: public-wai-ert-request@w3.org On Behalf Of Johannes Koch Sent: Wednesday, March 30, 2005 16:54 To: public-wai-ert@w3.org Subject: What is a subject of a test? When implementing EARL I asked myself "what is the subject of a test" and "how many asertions"? Example: HTML test document with two images. Test: look for alt attributes. Result: None has alt attribute. Version 1: One Assertion, one subject (URL of document), one testCase (alt attribute test), one result (fail) with two messages (about each of the images, including location information) Version 2: Two Assertions, each with same subject (URL of document), same testCase, one message each (including location information) Version 3: Two Assertions, different subject (URL of document + xpointer of error instance), same testCase, one message each Comments? -- Johannes Koch - Competence Center BIKA Fraunhofer Institute for Applied Information Technology (FIT.LIFE) Schloss Birlinghoven, D-53757 Sankt Augustin, Germany Phone: +49-2241-142628
Received on Wednesday, 30 March 2005 20:38:15 UTC