Ladislav Thon

The first question is: what priority order should we use? The CDI specification generally orders things in the ascending manner (interceptors, decorators, observers), but there is at least one existing...

> I think [Weld does that as well](https://github.com/weld/api/blob/master/weld/src/main/java/org/jboss/weld/inject/WeldInstance.java#L114-L143), so make it two :) Ah, totally forgot. Thanks! > We would also need to define behavior for beans that have no...

> Since SmallRye is looking for a specific method with specific parameters I think it would be better to use `Class.getMethod(...)` on each class/interface instead of getting all methods and...

Agree. Note that `Test*` is by default also a test class in Surefire, so Surefire will scan `TestClassMaker` for test methods :-) But it will find none, so it's OK.

Agree this doesn't make much sense for other scopes than `@Signleton` and `@ApplicationScoped`. For `@Dependent`, the solution is simple, just create an instance and immediately destroy it; for other scopes,...

> I've seen support tickets where the problem turned out come from calling toString() on CDI beans when writing out trace. So I'd recommend against this. calling `equals(null)` would be...

We unfortunately don't have the luxury of adding a `@Startup` annotation, because we already added the `Startup` event. Hence, the proliferation of ideas coming with the name `@Eager`. If we...

Well clearly many specifications (it's not just CDI, but also [REST](https://github.com/jakartaee/rest/issues/1262) and others) didn't get the memo. I know I didn't. Communication is hard. Apparently, the REST people agree with...

My experience with updating the TCK audit files is that the current CDI TCK model doesn't work well enough. I've explored one option (see #259), which doesn't work well enough...

Yeah, but the original model also required (and still requires) creating a second [near] copy of the entire specification, which is quite bad (unless we have a person fully dedicated...