On Wed, Jun 9, 2010 at 12:44 PM, Simon Laws <[email protected]> wrote: >> >> Right now it only really works well when using the Hazelcast based SCA >> binding (ie by including tuscany-binding-hazelcast-runtime). That uses >> the configuration that is being used for the endpoint registry so the >> SCA binding "just works". The problem with using other bindings for >> the SCA binding is in some environments we don't know what the >> absolute service endpoint is, eg when using web services and the >> webapp servlet host (and this was one of the main reasons for doing >> the hazelcast binding). >> >> ...ant >> > > OK so you're saying the only one that works properly at the moment is > the hazelcast based one. Assuming that the other ones worked though I > think the exiting tests cases for rmi and ws alternatives are wrong in > that the uris are specific. I would expect the ws and rmi variants to > "just work" in the same way as hazelcast, i.e. based on the endpoint > information available in the registry, so there is no requirement for > the user to specify this information in the composite file. >
Yep i agree with what you're saying about the existing tests, and even with other non-SCA bindings too you shouldn't generally specify an absolute uri on a service. I think there are likely problems with the endpoint information that goes into the registry for the different bindings, most of my testing has just been with the hazelcast based sca binding, the others aren't consistent and use different base addresses that may or may not be accessible across the domain. ...ant ...ant
