Change and compatibility rules for config=false (state) data

[BalazsLengyel]

Allowed changes and what is compatible should be specified differently for config=true and config=false data.

E.g. making an optional leaf mandatory is non-backwards compatible for configuration data and thus not allowed in RFC7950. The same change for config=false should be a compatible change.

[abierman]

What did the Version design team decide about this issue? Seems like an old-client will expect this data if madatory=true. If changed to false then the expected data is missing from the retrieval reply

More Discuss

[janlindblad]

I agree the rules should be different for config true and config false. For config true data, it's BC to make something optional, but for config false, the opposite is true.

[reshad-equinix]

I just got PTSD (from version design team discussions) seeing this issue

[kwatsen]

Seems reasonable to clarify opposite behavior for config false, output, and notification