The committers listed above are authorized under a signed CLA.

• ✅ login: zeitlinger / name: Gregor Zeitlinger (02b652e, 1067dee, 160ff44, 1c0f03a, 1c3dd54, 1fa6ddc, 34e2c75, 3729c7c, 419eddf, 4657008, 5420b3d, 553a041, 6212e6d, 70e012c, 7f2e9d2, 83715f9, 84a7dc0, 869fded, 8964862, 8c30ab4, 8cb016d, 909c6f7, 99cce85, a67aef8, ae26805, b0867a8, b1a25a4, b7f5ec4, c63e51a, cd3293a, cd9a76d, d2d445e, d47a380, e9c39af, eb18a06, f5f8bd0, f6d56f4, ffde54c)

@zeitlinger can you find ways to split this up into smaller pieces?

@zeitlinger can you find ways to split this up into smaller pieces?

good question - it's mostly tests, but adding those later will only reduce test coverage in the meantime. Do you have an idea?

you can see how I've broken out the database semconv stabilization tasks in #12608 in case that helps

you can see how I've broken out the database semconv stabilization tasks in #12608 in case that helps

you split off the tests - so I'll also do that: #12601

can you split the semconv changes up instead of implementing all of the changes in one PR?

you split off the tests

yeah, this sounds like a good option, then you could even split the tests and send only grpc test first for example

For reference, trask asked in #16130 how isPredefined would be implemented. Here's a gRPC example showing all three new methods:

@Override
public String getRpcMethod(GrpcRequest request) {
  return request.getMethod().getFullMethodName();
}

@Override
public boolean isPredefined(GrpcRequest request) {
  return true;
}

@Override
@Nullable
public String getErrorType(
    GrpcRequest request, @Nullable Status response, @Nullable Throwable error) {
  if (response != null && response.getCode() != Status.Code.OK) {
    return response.getCode().name();
  }
  return null;
}

For gRPC, isPredefined is trivially true since all methods come from protobuf service definitions. The distinction matters more for frameworks where arbitrary/dynamic method names are possible — the default false means "use _OTHER".

For gRPC, isPredefined is trivially true since all methods come from protobuf service definitions. The distinction matters more for frameworks where arbitrary/dynamic method names are possible — the default false means "use _OTHER".

I'm not sure this is the whole story for gRPC, see #15690

I can help to send a PR for apache dubbo after PR2 is merged.

@steverao PR2 is merged now: #16304 (comment)

Thanks @steverao for landing the dubbo PR! 🎉

Updated the table — PRs 1-3 are all merged now. Remaining: aws-sdk (25 files), rmi (4 files), gwt (2 files). All depend on grpc which is merged, so they can be created anytime.

Would you be interested in picking up rmi and/or gwt? Both are small — rmi is ~4 files (rmi + spring-rmi-4.0), gwt is ~2 files (gwt-2.0). I can provide pointers to the patterns from the grpc PR if helpful.

Thanks @steverao for landing the dubbo PR! 🎉

Updated the table — PRs 1-3 are all merged now. Remaining: aws-sdk (25 files), rmi (4 files), gwt (2 files). All depend on grpc which is merged, so they can be created anytime.

Would you be interested in picking up rmi and/or gwt? Both are small — rmi is ~4 files (rmi + spring-rmi-4.0), gwt is ~2 files (gwt-2.0). I can provide pointers to the patterns from the grpc PR if helpful.

Sorry for delaying response, I just saw it just now, recently, I will develop dubbo related prototype PRs about dubbo firstly. I can do them after dubbo.

Closing — RPC stable opt-in work has been delegated elsewhere.