Ondrej Lehecka

I would say METADATA_PUSH should be included in the resumable buffers. Its much easier of a mental model to think that no frames are lost. We can adjust the course...

- [ ] make sure stream state machines dont call subscription::cancel after on{Complete,Error} - [ ] make sure stream state machines dont call subscriber::on{Complete,Error} after cancel()

- [ ] cleanup Payload API - [ ] make sure that writing a payload to a stream from the right thread is going to synchronously be written all the...

Speaking about C++ implementation.. I can see how we can modify the RSocket library to allow emitting REQUEST_CHANNEL frame immediately followed by several PAYLOAD frames. On the responder side the...

we should remove the GFLAG from the library and let the application code specify the version via SetupConnectionPayload.

I used the next_adapt_stream implementation as a template for this algorithm. I can see how both next_adapt_stream and complete_adapt_stream can be written on top of adapt_stream. I am assuming this...