- From: Jose Kahan <jose.kahan@w3.org>
- Date: Thu, 8 Jul 2004 18:57:24 +0200
- To: www-xkms@w3.org
- Cc: dom@w3.org
Hi, While talking with Guillermo today, we found the problem of how we are going to test multiple clients and servers: - Should each server be tested by all clients and viceversa? - What happens when clients and servers only implement parts of the spec? - How can you make sense of the results, claim that your client or server are interoperable? I talked with the QA people and got the following pointers: - Organize your scenarios into profiles. - Have a reference client and server for each profile (meaning that they implement correctly everything in that scenario) - Make all the clients tests against the ref. server. - Make all the server tests against the ref. client. - Declare in the test suite which is the ref. client and server for each set of profiles. - There can be more than one reference client/server per profile. - If the client/server works against the ref. version, then it has passed the test. Does this make sense to you? BTW, I also got a pointer to some guidelines for writing test scenarios[1] and related info [2]. I asked again to see if we can have someone from the QA-WG attend the next XKMS teleconf. -jose [1] http://www.w3.org/TR/2004/WD-qaframe-test-20040225/ [2] http://esw.w3.org/topic/QA
Received on Thursday, 8 July 2004 12:57:32 UTC