Rich 'Shirley' Scheuerle wrote: > > With dim's changes Axis now passes the javax_xml_rpc.Service tests. > However, the change that allows GetPortsPosTest2 to pass is not JSR > 101 compliant. (See below..) > > So the question is "Should we change Axis to pass the TCK even if the > change is not spec compliant ?"
I would say that spec compliance is more important than passing the TCK.
Glen Daniels wrote: > > Hm. My personal preference is to leave our code be in places where > we think we're doing the right thing and the spec/TCK might be wrong, > and take it up with the Sun guys. The failed test will stand as a > reminder of the issue, and we can fix it later depending on the > resolution. I'd rather that than passing the test and then > forgetting about the issue or later saying "well what's done is > done", especially in cases where the effort to pass the test is a bit > more than this example here.
I agree. Note: we are not going to hear much from Sun either way for a week as the entire corporation is on mandatory vacation, but immediately upon their return, we can set up a conference call with Lance and work through the issues.
My recommendation is that we start numbering our issues (like Rich has been doing) and then annotate our TCK summaries with these numbers for every contested failure.
-Sam Ruby
