Steve: Developing on the Edge - feedback
Steve: Developing on the Edge
Thoughts on development, Web-services, technology and mountains.
7Jul
Fri2006
feedback

I don't want this to escalate, so will point people at Jorgen's rebuttal and say, yes, he may have a point.I'm pleased to see interop tests between arbitrary stacks will be allowed at the plug fest, and, best of all, WS-Addressing now has a test suite.

Note that the xmlns of WS-A is 2005/08, and that if you check the ws-a-test mail list, up to 2005/08, exactly one test had been proposed. One of the problems with doing tests after the rest of the spec has been done is that it makes it too easy to include stuff in a spec that isn't testable, and you are reduced to identifying which bits can be tested, and which bits cannot

We hit this problem in the CDDLM deploy api. The test plan includes some stuff that aint possible to generate from a correctly functioning SOAP stack just to see what happens, and, if you look at the results, it's clear that the spec includes stuff we can't test, which was a mistake. Optional things are the other troublespot too. In the analysis, someone has been fairly harsh about some bits of the design (i.e. my work) and the limited aspects of the SmartFrog implementation (my work again). I wrote those paragraphs. I'm not above critiquing my own process alongside that of the other specs I depend upon. And because I'm being harsh over our own process, it is only fair that people up the specification chain get the same treatment. Sorry.

I also need insert a disclaimer here. Text at this URL doesn't reliably represent my own opinions, let alone any other organisation. This is not 'someone at hp criticising MS', just someone who has written a half-completed SOAP stack in his spare time complaining about how hard it is to get it to work. Last year I blamed O/X mapping, this year, lacking that excuse, I have to pick on WS-Addressing and WSRF. Bit of a trend there. I guess if I switch to REST I will have to go after XML and HTTP1.1.

Finally, PJR has a new doc up on SOAP services from NetKernel, with some opinions on the specs too.. More negative than mine I think; I actually quite like SOAPFaults. We need to have a local interop event I think, perhaps at the highbury vaults. All SOAP stacks welcome.

Comments