Version: 2 Date: 2023/04/06
The current state of software development requires serving information to multiple consumer applications. Nevertheless, the base technology of the serving side does not support the transformation of information to the different requirements of the consuming sides.
There are existing technologies that solve this need, nevertheless, they have breaking changes and non-standard usage of the existing protocols, which has side effects like increased complexity of caching systems.
Alternatively, architectural patterns have also emerged to solve the need, but they increase the cognitive load by abstracting transformation in architectural pieces and increase the networking latency of the response as they behave as a proxy service.
- Allow filtering of the response on data retrieval.
- Follow current rest specification
- Keep compatibility with the current cache capabilities.
- Not intervene in the data sourcing process.
- work over JSON responses.
- Not increase network latency.
- Keep it simple to understand.
To achieve the goals expressed before the taken approach is the use of a query language based defined as a subset of the existing gql that will be passed in the requests of the consumer application as a query parameter field called fql of the standard GET requests.
The producer will interpret this parameter and intercept the response and filter the fields requested in the fql query parameter.
For the response the client will recieve, the next cases are posible:
- all fql gql parameters are available in the response: the server will return a 200 (OK) status code and the body will contain the filtered response body.
- A subset or none of the fql gql parameters are available: the server will return a 206 (Partial Conten) status code and the body will contain the filtered response body.