Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 6 Next »

####TODO: Peter Monks (and anyone else who's interested!) to flesh this out based on prior experience.  Topics may include:

  • Summary of current Symphony's versioning approach:
    • YAML file version is pinned to the current version of Symphony (e.g. 1.45, 1.46, etc.), and this mostly doesn't matter since the YAML file version isn't used for anything.
    • Each API endpoint independently has a version (v1, v2, etc.)
    • Currently, the Symphony engineers roll the YAML file's version for every release of Symphony.  They also roll an API endpoint's version any time anything changes in that endpoint - a conservative (safe), but high effort approach.
  • Endpoint Versioning vs API Suite Versioning - which is better?
    • Many developer platforms seem to have standardised on API suite versioning, so that approach would adhere to the Principle of Least Surprise (as one of many potential comparative criteria).
    • But the current Symphony API versioning approach doesn't do that, so there's a grandfathering problem (another potential comparative criteria).
    • Best Practice from BNY Mellon - is that MAJOR version in the semantic versioning scheme is sufficient. The major version is part of the REST API base URI. 
  • When do versions need to change?
    • That's easy: at least when a backwards compatibility breaking change is made (Symphony's current API versioning policy exceeds this requirement).
    • Whenever backwards-incompatible API changes are introduced, the version number should be incremented.

    • Conversely, for backward-compatible changes, do not change the version number but instead include a revision header. 
    • So what constitutes a backwards compatibility breaking change?
      • ####TODO
  • Version numbers as communicators of change:
    • Semantic Versioning (semver) clearly identifies how version numbers should be used to communicate change, and is already mandated for use  elsewhere in  the Foundation.
    • However, in the API realm, consumers generally don't care about backwards compatible changes (since those don't "ripple out" to impact their code), so is semver appropriate here?
    • Using version numbers in URLs forces incompatible changes to be communicated to clients and forces clients to make required changes before using the new API.  
Example: https://apigateway.example.com/positions/v2/holdings
  • APIs that return meta information (available API versions on this particular pod, build numbers to help with support requests, etc.)
  • As new versions of an API are released, at least one previous version should be maintained. An old version that is still being supported can be explicitly called as in https://apigateway.example.com/positions/v1/holdings



  • No labels