Question 31
Choose 1
An organization has created an API-led architecture that uses various API layers to
integrate mobile clients with a backend system. The backend system consists of a number
of specialized components and can be accessed via a REST API. The process and
experience APIs share the same bounded-context model that is different from the backend
data model. What additional canonical models, bounded-context models, or anti-corruption
layers are best added to this architecture to help process data consumed from the backend
system?
Correct answer:
C. Create a bounded-context model for the system layer to closely match the backend data
model, and add an anti-corruption layer to let the different bounded contexts cooperate
across the system and process layers
Explanation:
Why C is correct
In API-led connectivity + DDD terms:
Your backend system already has its own data model (and it’s accessed via a REST API).
Your process + experience APIs intentionally share a different bounded-context model (a consumer/business-oriented representation).
The clean, recommended way to connect those without polluting the upstream model is:
System API bounded context ≈ backend model
The System API is meant to “encapsulate the system of record” and therefore typically aligns closely to the backend’s model (so it can expose that system consistently and avoid forcing the backend to conform to upstream semantics).
Anti-corruption layer (ACL) between system and process
The ACL performs translation/mapping between the backend/system model and the process/experience bounded context, preventing the backend model from “leaking” into the upstream domain model (and vice versa). This is exactly what an anti-corruption layer is for: enabling cooperation across bounded contexts while preserving boundaries.
Why the other options are worse
A: Having a different bounded context per layer can be valid, but “overlap them” and “let developers know about differences” is basically accepting model leakage and inconsistency rather than containing it with an explicit translation boundary.
B: A single “combined canonical model” that merges backend + API-led models is a classic trap: it tends to become a lowest-common-denominator model that fits nobody well and increases coupling. It doesn’t respect bounded contexts.
D: “ACL for every API to match each other” creates N×M transformations and pushes you toward a fragile, hard-to-govern mesh of mappings. You typically want one deliberate boundary translation where models change, not everywhere.
Bottom line:
Keep the system layer close to the backend, and use an anti-corruption layer to translate into the process/experience bounded context.
Why C is correct
In API-led connectivity + DDD terms:
Your backend system already has its own data model (and it’s accessed via a REST API).
Your process + experience APIs intentionally share a different bounded-context model (a consumer/business-oriented representation).
The clean, recommended way to connect those without polluting the upstream model is:
System API bounded context ≈ backend model
The System API is meant to “encapsulate the system of record” and therefore typically aligns closely to the backend’s model (so it can expose that system consistently and avoid forcing the backend to conform to upstream semantics).
Anti-corruption layer (ACL) between system and process
The ACL performs translation/mapping between the backend/system model and the process/experience bounded context, preventing the backend model from “leaking” into the upstream domain model (and vice versa). This is exactly what an anti-corruption layer is for: enabling cooperation across bounded contexts while preserving boundaries.
Why the other options are worse
A: Having a different bounded context per layer can be valid, but “overlap them” and “let developers know about differences” is basically accepting model leakage and inconsistency rather than containing it with an explicit translation boundary.
B: A single “combined canonical model” that merges backend + API-led models is a classic trap: it tends to become a lowest-common-denominator model that fits nobody well and increases coupling. It doesn’t respect bounded contexts.
D: “ACL for every API to match each other” creates N×M transformations and pushes you toward a fragile, hard-to-govern mesh of mappings. You typically want one deliberate boundary translation where models change, not everywhere.
Bottom line:
Keep the system layer close to the backend, and use an anti-corruption layer to translate into the process/experience bounded context.
