Thanks for finding that Simon. I think when I last came across that I parsed it as assuming a local, PBR interface. But...I don't have any support for that and the surrounding text specifically mentions serialization, so I think you've captured the intent.
Well, it still seems ugly to me... if something in our Tuscany impl changes, then your WSDL-mapped interface is now different? It seems like we're talking about a class of service: remotable in the sense of cross-JVM, but never invocable outside of the Tuscany runtime, not very SOA-ish. I'm not saying it's useless... it's just kind of a new concept thrown in from my perspective. Anyway, I opened https://issues.apache.org/jira/browse/TUSCANY-3911 as it doesn't work currently, so when someone gets a chance we can resume this discussion. Scott On Thu, Jul 28, 2011 at 6:12 PM, Simon Nash <n...@apache.org> wrote: > Scott Kurz wrote: > >> Simon, >> >> I didn't find the spec requirement to support this, do you happen to >> see it? BTW, as I mentioned in TUSCANY-3894, I did leave the >> current behavior so that it preferes Java serialization over the local >> binding-sca-runtime path. >> >> Ths, >> Scott >> >> >> From section 7.2.5 in the latest javacaa spec version that I have: > > ServiceReference objects can also be passed as parameters on service > invocations, enabling the responsibility for making the callback to be > delegated to another service. > > Although this is mentioned in the context of callbacks, the statement > in the first part of the sentence isn't limited or qualified as only > applying to this particular usage. > > I don't think it matters whether Tuscany uses Java serialization or > XML serialization when passing service references, because Tuscany's > ServiceReferenceImpl supports both of these with identical semantics. > > Simon > >
