HTTP/2 support and Cloud Native Microservices are definitely two important goals we should achieve in this year, and I am looking forward to reading your proposal. Meanwhile, I think we should discuss and come up with a draft roadmap for the incoming whole year so that it will make the community easier to participate.
What do you think, Jun? Thanks, -Ian. On Thu, Apr 18, 2019 at 9:46 PM Jun Liu <[email protected]> wrote: > Hi, All > > I am writing this mail to discuss about our work in the next stage, some > trends we should follow, or even what Dubbo should be next. > > Approximately from last year, there have been some discussions and > thoughts in the community about how Dubbo should following or leading the > tech trend in its area. With this goal in mind, I recently spent some time > trying to dive deep in some related techs. Here I want to express some of > my points regarding the two most discussed topics in our community, they > are, HTTP/2 protocol and Cloud Native. > > 1. http/2. Considering that gRPC has done a great job upon HTTP/2 and the > extraordinary influence it achieved in the industry. I think it should be a > non-negligible part for us to take into consideration when talking about > this topic. Actually, I would suggest we first consider integrating gRPC > into our framework to 1. to leverage Dubbo's more powerful service > management ability 2. quickly bring Dubbo HTTP/2 support. This is suitable > to be done in 2.7.x. In the long run, I think we may need to provide > built-in HTTP/2 support, which may require a standard HTTP/2 application > layer protocol (compatible with GRPC can be an option), and introduce > application level streaming semantics. This is suitable to be done in 3.x. > > 2. Microservices and Cloud Native. When comes to this topic, the most > obvious thing to notice is the mismatch of the service model between Dubbo > and Microservices architecture. Our definition of services is based on > Interface, one Interface is intercepted as a service, while the > Microservices architecture, Application is the first-class citizen, you > can consider an Application as a service, interfaces are only see as > metadas inside the service. This mismatch makes service discovery, service > management and any introps very difficult or even impossible. I like the > Microservice's model more than Dubbo's and I think we should start > reconsidering our service model. I would restrict the discussion about > Cloud Native to K8S here, I continuely see issue asking about Dubbo running > on K8S. I think this question on root is about how to leverage the service > discovery and container orchestration capabilities provided by K8S. > > This time I think we should not just stay in discussion, instead, we need > to go to the next phase to come up with some practical solutions, make them > part of our milestone and realize them. If you agree with the opinions I > put forward above, I am glad to try to write down some of my recent > researchs into a draft proposal. Then we can further discuss on some of the > details and finally make it a practical one. > > Ken > >
