Consider changing the enum type under Server Variable

[webron]

Right now the type is an array of primitive. Variables are not typed, we were just looking to restrict values to non-complex objects.

We can change it to string - which would simplify tooling. This will affect how the following example would be though:

    port:
      enum:
        - 8443
        - 443
      default: 8443

That would become:

    port:
      enum:
        - '8443'
        - '443'
      default: '8443'

which is a bit more annoying to write.

Changing the type of enum would also change the type of default.

[ePaul]

Would this mean that the type: string is superfluous for a schema definition with enum, as other types would not be possible anyways?

[webron]

Not sure I follow. I'm talking specifically about https://github.com/OAI/OpenAPI-Specification/blob/OpenAPI.next/versions/3.0.md#serverVariableEnum.

[DavidBiesack]

I support changing from primitive to string. Under Server Object, variables is defined as "An object holding variables for substitution in the URL template." We should not expect/suggest OAS users impose semantic types like "ports are integers" -- this is not a port number, it a piece of a larger URL string.

While you are in there, I suggest

1. Update the text in Server Obejct to say "An object holding string variables for substitution in the URL template."
2. Update "will be sent if " to "will be used if" (we don't "send" anything here.)
3. Please fix the markup typo OpenAPI Object's servers(#oasServers):
4. Please add doc clarifying how/where/when these variables are specified and when the substitution occurs. (Other uses of 'consumer' in the spec refer to the service consuming a request body, or users of the API service, so 'consumer' here may be confusing. Since I'm not clear when/how these parameters are used, I can't suggest another wording.)

[ePaul]

@webron Sorry, I misunderstood this somehow thinking that you want to change the meaning of the enum keyword.

You are right, what we are doing here is describing a string variable (as @DavidBiesack elaborated), so the enum should also consist of strings.

[darrelmiller]

@DavidBiesack "How/where/when" these variables are specified will depend on the purpose and implementation of tooling. I don't believe this can be specified in the spec. I also don't think it needs to be.

[DavidBiesack]

ok, but I still don't know what these variables are really for. I don't mean that we specify, only that we describe (perhaps in Server Object Example) some scenarios where the variables may be used... why these exist.