On Thu, Apr 23, 2009 at 3:51 PM, Scott Kurz <[email protected]> wrote: > On Wed, Apr 22, 2009 at 3:55 PM, Raymond Feng <[email protected]> wrote: >> 2892 If a reference does not have a binding, then the binding used can be >> any of the bindings >> 2893 specified by the service provider, as long as the intents attached to >> the reference and >> 2894 the service are all honoured. > > That's interesting. So this is the OASIS Assembly Spec Sec. 8.5. > > I see a number of problems with this section: > > 1. By "reference does not have a binding" it is not clear if they are > talking about the lack of a binding element (in which case the > <binding.sca> might be set to be implicitly present) or the lack of a > binding, period (in the case of this SCAClient-created proxy), or > both. I'm guessing they mean both. This is a significant change > from OSOA. But this is only a minor clarification needed.
I'm not an expert but as I understand it it means that when there is no binding element explicitly provided on the reference then the binding is selected from the bindings listed on the services. Note also the following... 904 • If a reference has a value specified for one or more target services in its @target attribute, 905 there MUST NOT be any child <binding/> elements declared for that reference. > > 2. > 2876 ..... If the interface of the service or reference is > remotable, then either > 2877 the local or remote variant of the SCA binding will be used > depending on whether source > 2878 and target are co-located or not. > > This is confusing. It almost sounds like for a co-located service > with a remotable interface, you would expect PBR instead of PBV. > Though there must be a local vs. remotable variant of the SCA binding > impl, there may or may not be a co-located vs. non-co-located variant. > This text should probably just be deleted. Yep, I don't see that the spec defines different semantics for binding.sca depending on whether a target is local or remote. Although there are likely to be implications for resource used, performance etc between the different versions of binding.sca. > > 3. organization - This specification of a behavior about references > shouldn't be buried in the "binding.sca" section of the spec. It is described in the sections on reference/service also. In particular there is a section "4.3.1 Specifying the Target Service(s) for a Reference" that describes it in some detail. > > I haven't personally been participating in the OASIS spec > discussions.. not sure if anyone is following this thread who is would > agree that this is worth bringing up... or I could jump in that OASIS > mailing list I guess. > > Scott >
