On Dec 12, 2007 3:29 PM, Mike Edwards <[EMAIL PROTECTED]> wrote: > Folks, > > A thought here is whether we should be keeping around all of the > original SCDL at all times. > > I can regard the Domain as consisting of the SCDL made up from all the > contributions made into it. I have the view that the Domain is a kind > of distributed database of configuration information. I think this is > separate from the runtime information within Tuscany which is an > implementation of the "design" laid down in the Domain configuration. > > So perhaps the idea should be to always keep the Domain information (XML > form if desired) and to reference it from the runtime entities. That > way, it will be always possible to go back and retrieve the original > Domain configuration information. > > Thoughts? > > Yours, Mike. > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > I think we should have it available. Primarily I think it allows for subsequent matching operations but this conversation has raised another question in my mind. It we serialize a composite with a component reference that was originally defined with multiple bindings but only one wired target then I believe at the moment we will just see the single binding in the output. We won't see the range of bindings that was originally declared. Should we see those bindings or not? If so how should they appear? I.e. should we present the reference bindings with complete URIs, where they have been completed, or should the be presented as they were originally described?
B.t.w. following on from Mikes point I'm not necessarily saying it has to be held inside the assembly model but it would certainly be convenient to have it there. For example, by holding a "candidateBindings" list or similar which starts off being the same as the bindings list read in from the reference definition. Simon
