Marty Humphrey wrote:
I don't like the sound of this.
For me, the real question is - if the existing tooling does not support some approach -- HOW DIFFICULT is it to manipulate the existing tooling (e.g., a small hunk of additional code, perhaps) and HOW SOON is this additional code projected to appear in the tooling itself (this projected date must be agreed-to via consensus)?
It seems overly restrictive to say that if Sun and/or Microsoft et. al. don't support it today, then we can't do it.
-- Marty
In the CDDLM implementations we demonstrated last week, we were 100% sure that we would not be able to demonstrate interoperability. How so? because the three different implementations of WS-A were all different -NET 1.1/WSE -Axis1.2.x+Apache Muse -Axis2/0.92 I ended up writing classes to try and map from the various different forms of WS-A that could be seen in messages to that of the stack, and had to give up on the reference paramaters bit because the underlying stack assumed they were string values. How can SOAP stacks get away with getting something as fundamental as the address wrong? Because there is not a single test for WS-A. Only now that the spec is nearly a release issue has a mailing list been set up, to which there is currently one whole proposed test: http://lists.w3.org/Archives/Public/public-ws-addressing-tests/2005Aug/0000.... How then can we blame the development teams when they dont get any way of verifying that they have got it right? It is not their fault, they have been given inadequate specifications with no means of verifying compliance with the specification. Similarly, the number of tests for WS-RF and WS-N appear to be less than that for WS-Addressing. Yet we are avocating it as the architecture for global scale, long-lived distributed systems. Marty (and anyone else) : if you are prepared to write the test cases for what constitutes a compliant implementation of WS-A, and get them accepted by the WS-A committe, I will personally integrate those tests into Axis2, which, by consequence of the no-ship-without-100%-tests-passing policy of the team, will then guarantee that Axis2 is compliant with the specification. Without those tests, all bets are off -you will have to take what you are given. Schedule-wise, Axis2 is at the 0.93 release, with a few more iterations before it ships at 1.0. Now is the best time to write those test documents if you want functional WS-A before the end of the year. I wouldnt worry about Sun's stack, and the MS stack is out of my control. Ask Savas about that one. -Steve