Stabilize opentelemetry-semconv

[jack-berg]

Want to start a conversation about what stands in the way of marking the opentelemetry-semconv artifact as stable. We've come a long way recently, re-organizing by domain, switching from build-tools to weaver based code generation, and changing how we think about deprecated / incubating attributes.

[breedx-splk]

With incubating in place, it seems like most of the roadblocks have been cleared. I suppose the main thing to think about is needing to be able to support future breaking changes though major semver bumps, which should be seldom and annoying at worst?

[jack-berg]

It would also be good to think through what might change in a future where we're generating helpers / constants for producing semantic convention metrics (#12), events, and anything else.

Does the class organization around root namespace work for this? For example, we currently have classes named HttpAttributes and HttpIncubatingAttributes. These names imply that the only hold the http attributes. If we added metric generation, would we feel comfortable adding classes like HttpMetrics, HttpIncubatingMetrics?

Maybe we think of the current class organization as the embodiment of the semantic convention attribute registry - a global registry of attributes agnostic of which conventions they're used in. Then, we can independently decide how to organize anything else we generate for specific conventions.