Can You Bypass the Mid‑Platform? When Direct Calls Across Service Layers Make Sense

After a company completed a mid‑platform (中台) architectural overhaul, it achieved good results but raised a new question: should calls strictly follow the layered design, or can they cross layers when appropriate?

1. Can the channel or scenario product layer call backend modules without going through the mid‑platform?

The answer is yes. For standardized services such as detail queries or shared utilities, direct calls are acceptable. Enforcing the black‑arrow, strict‑layer path only adds extra nodes to the call chain, increasing latency without providing governance benefits.

The only effect is that the scenario product team no longer needs to adapt to the technical standards of each core service module; the mid‑platform assumes that responsibility. This mirrors the DDD‑style COLA framework, where the application layer can invoke the infrastructure layer directly for simple query interfaces.

2. Can lower‑level backend systems call upward without violating the Dependency Inversion Principle?

Again, the answer is yes. In banking, core systems must push real‑time settlement notifications upward to reach customers, and they may need to forward transaction data to higher‑level product services for business‑rule evaluation. Some upward dependencies are inevitable and do not constitute a design flaw.

3. What is the subtle difference between “call” and “push” in service chains?

When discussing service interactions, the author distinguishes top‑down "calls"—used to satisfy specific business requirements by invoking dependent services—from bottom‑up "push" operations, which are typically notification‑type transactions with minimal business coupling (except for cases like inbound transfers that require rule checks).

At present, there is no widely accepted terminology to capture this nuance, but recognizing the distinction helps architects evolve platform designs more thoughtfully.