On Wed, Jan 20, 2010 at 6:30 PM, Raymond Feng <[email protected]> wrote: > Hi, > > In the distributed environment, when the client tries to talk to remote > service, it could be some latency before the service is available. For > example, it takes time for the endpoint registry to be updated with a remote > service. > > In general, there are a few typical options to handle such latency. > > 1) The client programming model has a timeout, for example, OSGi > ServiceTracker.getService(timeout). SCA doesn't have such concept though. > Potentially, we can use some intents to express that. > > 2) The service proxy retries for a few times. For example, Axis2. I also saw > some code in our endpoint registry that retries for a few times when the > target endpoint cannot be found. > > Any thoughts? >
Is this just about in Tuscany tests where multiple Nodes are used and one Node uses services in another Node so needs to first wait for the other Node to finish being started? That seems like a slightly contrived situation that wouldn't be so common in real world use, so unless there's more to it then it seems simplest to not try to do anything - if an endpoint isn't found just fail the request, if a test uses multiple Nodes then it can add Sleeps itself to give time for things to start up. ...ant
