Why is Hongmeng's soft bus a wrong design?

The soft bus is just a name, and there is nothing to criticize about it. For Taiwanese, it might be called bus flow.

To say that the soft bus is a wrong design means that it has returned to RPC.

RPC is a stateful thing, and it has operational semantics. This degree of freedom is too large and there is no good abstraction. This directly leads to only end-to-end between the two points ab, and this degree of abstraction is not enough.

Can't b be a group of services? Can't it be aggregated like the facade model? Can a/b be symmetrical? Reuse a transport?

These are broader scenarios. In fact, there is a historical reason that HTTP is not pure enough, but restful is undoubtedly correct (referring to resource model and crud); while mqtt is on the other side, it is resource semantics that implements the observer mode, and the important thing is that it does It is also a stateful thing, but state is a resource, and the server can be scaled.

The architecture and communication capabilities of a system determine how far the system can go. The designer of Hongmeng is just a vague idea. It can be based on wan to realize the design concept of distributed OS (in the 1990s), but the designer has not designed it. The system-level communication protocol does not know what the design selection principle of the system-level communication protocol (such as 9P) is. Of course, there will be no correct design without accurate vision.

The RPC design using end-to-end operational semantics has restricted this system to being unable to cope with future scenarios. When the number of nodes is large, the difficulty of collaboration increases geometrically. This is something 20 years ago, if RPC is correct, Jini would have ruled the world a long time ago.