Classification: UNCLASSIFIED Caveats: NONE Responses follow below.
Gang ---cut--- > To me this naming convention adds complexity and the possibility > for problems if the convention isn't followed. If a new type is > defined rather than using Map<String, Object> directly, this solves > the injection type problem, but it means that all application code > that uses this capability must add a dependency on a Tuscany-specific > class, which means that the component can't be run (or even loaded) > on other SCA implementations. Actually this is a missing piece in SCA spec in my opinion and I'm surprised. ---cut--- > Thanks for the clarification. If handlers want to share information, > they can create shared context (not the same as the "application > context") > to do this. I think it's undesirable to overload the purpose of the > "application context" by also using it to share state between handlers. > This exposes things to the application that are intended for handler > use > rather than for application use, which isn't desirable IMO. The purpose of "application context" is still sharing states. The argument is how much policing (restriction) you want to put. I'm not advocating for no-policing, but it has to allow doing simple things easily. Trade off is the usability. > > Simon > > > --- cut --- > > > > Gang > > > > Classification: UNCLASSIFIED > > Caveats: NONE > > > > > > > > Classification: UNCLASSIFIED Caveats: NONE
