open-feature / spec

OpenFeature specification
https://openfeature.dev
Apache License 2.0
595 stars 35 forks source link

Spec could be more clear about named-client/provider binding. Would "namespace" help? #195

Closed toddbaert closed 10 months ago

toddbaert commented 1 year ago

Named client / provider mapping is a useful and important concept. I also think the spec only does the bare minimum in explaining it. I'd like to propose two things:

Particularly interested in your thoughts on the namespace proposal @justinabrahms . I think regardless we can add more non-normative text.

cc @thomaspoignant @kinyoklion @beeme1mr @lukas-reining @benjiro

justinabrahms commented 1 year ago

As long as we change the signature to api.getClient(String namespace), then it seems fine to me?

toddbaert commented 1 year ago

As long as we change the signature to api.getClient(String namespace), then it seems fine to me?

Ya, I think we'd basically do that everywhere: name -> namespace as long as there's a general agreement this is actually an improvement.

Kavindu-Dodan commented 1 year ago

I think scope is also a good candidate. For example, a scoped client or a scoped provider where scope is a string . But I am also fine with namespace

kinyoklion commented 1 year ago

I agree that something other than name is good. I do have a bit of a concern about namespace/scope, which is when I hear namespace in this context I think of something like namespace:flagKey (routing:useV2).

Probably doesn't matter, as long as the non-normative text is illustrative enough.

toddbaert commented 1 year ago

I agree that something other than name is good. I do have a bit of a concern about namespace/scope, which is when I hear namespace in this context I think of something like namespace:flagKey (routing:useV2).

hmm, I agree.

toddbaert commented 10 months ago

Closing this for now. This seems to be well-enough understood.