Support @Stereotype

[t1]

The FT spec doesn't mention Stereotypes, so there is no test in the TCK and implementations don't have to support it.

It would be useful to, e.g., define a @Stereotype-annotation @Gateway with the complete default FT handling and use it for all gateway classes.

As a fallback it should at least be explicitly documented as not supported.

[t1]

@dblevins: I know you're a great supporter for meta-types. Don't you think requiring support for Stereotypes would be a good addition to fault-tolerance annotations? I'd like to have some support before I start to dig into this issue.

[Emily-Jiang]

I will defer this one. However, I will add a statement on the spec to say "The stereotype support is optional."

[Emily-Jiang]

We discussed this in today's hangout, here is what we found.

Stereotype is used on beans or method producer or field producer, not the aggregation of interceptor binding.

[t1]

@Emily-Jiang: I don't understand what this means, I guess it would have been beneficial, if I was on the hangout :-( It would be very kind of you, if you could expand a little on the topic?

[Emily-Jiang]

Stereotype is only used for CDI beans, producer fields and producer methods. Unfortunately, you are asking for a metatype aggregation of interceptor bindings, which CDI cannot do. Stereotype cannot help in this circumstance.

[t1]

Maybe I'm getting something wrong, but CDI Stereotypes can be used for Interceptors. See http://docs.jboss.org/cdi/spec/2.0/cdi-spec.html#specify_stereotype_interceptor_bindings

[Emily-Jiang]

Stereotype has scope, interceptors, qualifier. Interceptors alone can make a Stereotype with a default scope of Dependent though no scope is mentioned. Stereotype is only used for producers.

[OndroMih]

I'd like to clarify what @Emily-Jiang wrote: Stereotypes can be used only on producers and classes but on methods. So there's a value in supporting stereotypes but only as default configuration for all methods of a CDI bean. It should be clarified that putting a stereotype on a method would be ignored - only specific interceptor annotations can be used for methods.

[t1]

Okay, now I understand. This limitation of the CDI spec regarding stereotypes does restrict its usefulness, even when it never occurred to me before. I guess the main use-case is to define stereotypes on the class level anyway. When I have a bunch of gateway classes to encapsulate the access to remote systems that I want to protect with FT, wouldn't it be useful to define the default FT behaviour in a @Gateway stereotype annotation? I could change anything at the class or at the method level, but the defaults are there. I could not remove something, e.g. the timeout, but I could change it to any value.

I think this would be very valuable. Don't you agree?