Add rules in design guideline about multi-endpoint API design & implementation requirement

[bigludo7]

Problem description In discussion #160 and #212 we have long thread of discussion about implementation of multi-endpoint API.

Several opinions are provided but we have not a clear decision/pattern on how we manage this.

1. Do we decide that an API must be fully implemented (all endpoints have to be able to provide 2xx response) in order to be "certified"* ?
2. (if 1. answer is yes) If we decide that an API must be fully implemented in order to be "certified" do we add a rule to enforce multi-yaml definition when distinct endpoint tackles independent data?
3. (if 1.answer is no) If we allow partial implementation do we mandate use of 501 for non implemented endpoint? As suggested in the design guideline and outcome of discussion #160

Additionally mention of Runtime Restriction API should be probably in the mix for partial implementation but as we cannot assume it will be used by all API provider & consumer.

(* as we have a GSMA certification but not CAMARA certification we can assume here GSMA certification but this point should be discussed with @MarkCornall and @trehman-gsma)

Expected action Get a decision on these questions once for all and write some rules about it in the design guideline document

Additional context @rartych please do not hesitate to provide comments if I wrongly translate commonalities discussion we had July 8th

[hdamker]

As of TSC Minutes https://wiki.camaraproject.org/display/CAM/2024-07-18+TSC+Minutes:

• Do we decide that an API must be fully implemented (all endpoints have to be able to provide 2xx response) in order to be "certified"* ?
  • Tendency to yes and split to have distinct yaml if independent resource [option 1+2]
  • We need an option for an optional endpoint when they are on the same resource (we cannot split-up):
  • We can add only 501 when, for a multi-endpoint working on a same resource, we have optional endpoint
  • Mandatory endpoints must not have 501 defined as a valid response

[gmuratk]

Is there a PR in progress for this issue?