Doug Fawley

> this implementation has the same drawback as the previous one. A single extremely large message at the start of the stream followed by small messages for the remainder of...

> I have added a test for the shared buffer pool. If you require any additional tests, please feel free to let me know. I think it would be useful...

@ejona86 @markdroth What is happening in grpc-java/c++ here? Do you support both versions? v1 was added to the grpc-proto repo 4 years ago, but we never picked it up in...

@buzzsurfr SGTM, I think we can go ahead and support both v1 and v1alpha.

I think we will need one primary implementation and one wrapper implementation that adapts v1alpha requests & responses into v1 requests & responses. I think we can use `proto.Marshal` +...

Sorry, one more thing: `NewServer` should continue returning a v1alpha server, and `NewServerV1` can return the v1 implementation.

IMO that's a fine implementation detail, but the important part is: we should make the `Register` behavior register both v1 and v1alpha for now (ultimately only v1 in the distant...

@yurishkuro you will need to work around this on your end. We will need to compile in v1alpha protos for a very long time (possibly even forever?) for compatibility with...

Ping @ZhenLian

Ping @ZhenLian