Hi Raymond ...snip > Basically, each Node can be configured with the following information: > > domain: The URI of the domain that the Node belongs. > domainRegistryURI: The URI that can be used to connect to a domain registry, > such vm://localhost, multicast://<address>:<port>?attr1=v1&attr=v2. > > When the Node starts, it looks up the DomainRegistryFactory (replacing > EndpointRegistryWrapper) which knows how to find an implementation of > EndpointRegistry that support the scheme such as "vm" or "multicast". An > instance is created for the given registry URI and domain URI and cached by > the combined key. I removed the endpoint-wrapper module.
Am very happy to see the wrapper as I didn't know what it was for ;-) > > I also refactored the runtimebuild code from the EndpointReferenceBuilder > into a new utility called EndpointReferenceBinder that can bind a EPR to the > endpoint in the endpoint registry that is passed in as an argument. Sounds OK but I'd still like to be able to call this binder from the builders for those references that are resolveable locally. The intention being to warn users of incorrect wiring where possible at build time rather than waiting until run time. I'm not saying your change makes this any more difficult but you removed an empty method that was reminding me to look at it. I'll try and get round to it. > > When I tried to bring up the tests in domain-node, I discovered some flaws > in our Monitor creation and usages. I'll start a different thread for that. Ok, this is to do with having a separate instance of the domain for each node instance now that the top level design has been changed to share the majority of infrastructure between nodes? I'll look for the new thread. Simon
