On Wed, Jul 18, 2018 at 1:05 PM, Dennis Kliban <dkli...@redhat.com> wrote:
> I was asked on IRC to state what problems the proposed changes are trying
> to address. There are two problems I see with the current OpenAPI 2.0
> schema Pulp's REST API provides.
>
> - The path parameters in the schema don't reflect the parameters our
> users should be using for identifying the resources available via REST API.
>
I'm not convinced that we should use hrefs as the sole identifiers for the
resources.
Here are the reasons I see that we use hrefs as identifiers in a REST API
context:
1. Each href provides full context into the resource it identifies.
When given a href you would know exactly which resource it is referencing
and would never run into the issue of: what is this {uuid} because you know
it is a 'repositories/{uuid}'
2. discoverability, you know exactly how to access resources from
hitting the root url (and in a webui can just click)
3. You would not need to construct urls from templates
But things are different if we look at it from a bindings/client context.
The difference is mainly due to how discoverability is done: in the REST
API context the user has little prior knowledge to what resources are
available, and how to access theses resoruces. But the bindings/client are
generated from the schema, which defines exactly how resources are
structured, and what the context of each {uuid} is.
1. Given an {uuid} the client/bindings knows exactly what resource this
{uuid} refers to. With hrefs there would be redundant information
pulp.repositories('repositories/{uuid}') (why do I need to specify
repositories twice?)
2. Discoverability is done with the schema which contains all the
information about available resources/endpoints
3. URL construction is done by the client, so the user would also never
need to do any url construction themselves (unless we continue to force
href only identifiers, in which case they might have to do some url
construction to pass as arguments)
I don't think hrefs and uuid identifiers are mutually exclusive. I propose
that we extend HyperlinkedRelatedFields to accept either href or uuid, and
map these HyperlinkedRelatedFields to each other in the schema with openapi
definition objects [0].
[0]
https://github.com/OAI/OpenAPI-Specification/blob/master/versions/2.0.md#responses-definitions-object
>
> - The path parameters don't have a description in the schema.
>
> +1 to updating the schema descriptions for these parameters
> Do others agree with these problem statements?
>
>
>
> On Wed, Jul 18, 2018 at 9:31 AM, Dennis Kliban <dkli...@redhat.com> wrote:
>
>> I am working on improving the OpenAPI 2.0 schema for Pulp 3. I would like
>> to get some input on the improvements I am proposing. The schema is used to
>> generate our REST API documentation as well as the bindings with
>> swagger-codegen.
>>
>> The docs generated from our current schema look something like this:
>>
>> GET /repositories/{repository_pk}/versions/{number}/content/
>> <https://docs.pulpproject.org/en/3.0/nightly/integration-guide/rest-api/index.html#get--repositories-repository_pk-versions-number-content->
>> Parameters:
>>
>> - *number* (*integer*) –
>> - *repository_pk* (*string*) –
>>
>> Status Codes:
>>
>> - 200 OK
>> <http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.2.1> –
>>
>>
>> Since Pulp identifies resources using their HREFs, I am proposing that
>> the schema produce documentation that states:
>>
>> GET /{repository_version_href}/content/
>> Parameters:
>>
>> - *repository_version_href* (string) – HREF for the repository version
>>
>> Status Codes:
>>
>> - 200 OK
>> <http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.2.1> –
>>
>>
>> Thoughts? Ideas? All feedback is welcome. Thank you!
>>
>
>
> _______________________________________________
> Pulp-dev mailing list
> Pulp-dev@redhat.com
> https://www.redhat.com/mailman/listinfo/pulp-dev
>
>
_______________________________________________
Pulp-dev mailing list
Pulp-dev@redhat.com
https://www.redhat.com/mailman/listinfo/pulp-dev
