Upgrade Framing to Spec 1.0

[rsocket]

Upgrade Framing to Spec 1.0

[NiteshKant]

@benjchristensen sure! We should target 0.5.x as that is where all development is happening.

\cc @stevegury @robertroeser

[benjchristensen]

Sounds good. We will keep you updated as we make progress. Thanks.

[benjchristensen]

We plan to start on this soon as the spec is nearing completion and we're working on reactivesocket-cpp right now.

[benjchristensen]

Is it essential for the code to remain backwards compatible (server taking older frames), or can it be a breaking change where both client and server deployments need to be upgraded?

I assume backwards compatibility is needed, but it would be easier and cleaner to break.

[robertroeser]

I think a clean break would be ok - no need to add tech debt right now.

[benjchristensen]

Great, that makes this far simpler. Thanks Robert.

[benjchristensen]

Since we are going to be in flux for a bit, we are going to do this on a separate branch while we submit, review, and merge all the diffs for framing. So we don't break everyone on 0.5.x during that time, I have created 1.0.x for all of that right now.

Any issues with us targeting 1.0 Release Candidate once we get the framing nailed down? Or would you prefer something like 0.6?

[benjchristensen]

This relates to rsocket/rsocket#143 where we are approaching v1.0 of the spec. We will call it 1.0 Final once we prove the implementation in both Java and C++.

The Java and C++ 1.0 versions will happen whenever they consider themselves ready, not related to us considering the spec final at 1.0

[robertroeser]

We should probably add back in the TCK before we start making release candidates? Maybe version 0.6 after the framing stuff is added?

[lexs]

https://github.com/ReactiveSocket/reactivesocket-java/tree/1.0.x now contains mostly an up-to-date impl of the 1.0 spec. The bit missing is handling of the NEXT bit of payloads, both sending it and not handing data of to onNext() if it's not present.

I'll spend the rest of this date writing tests to verify that the wire format is according to spec, but I agree having the TCK would be nice before we release anything. I believe somebody from Facebook was going to start working on the TCK, @benjchristensen ?

[lexs]

@somasun filed #244 for TCK.

[yschimke]

Can this be closed?