Open Stack

Hi David,

Thanks for pointing it up.

> -----Original Message-----
> From: David Kranz [mailto:dkranz at redhat.com]
> Sent: Saturday, April 19, 2014 4:16 AM
> To: OpenStack Development Mailing List
> Subject: [openstack-dev] [nova][qa] Compatibility of extra values returned in json dicts and headers
> 
> Recently, as a result of the nova 2.1/3.0 discussion, tempest has been
> adding validation of the json dictionaries and headers returned by nova
> api calls. This is done by specifying json schema for these values.
> As proposed, these schema do not specify "additionalProperties": False,
> which means that if a header is added or a new key is added to a returned
> dict, the tempest test will not fail. The current api change guidelines
> say this:
> 
> Generally Considered OK
> * The change is the only way to fix a security bug
> * Fixing a bug so that a request which resulted in an error response
>   before is now successful
> * Adding a new response header
> * Changing an error response code to be more accurate
> * OK when conditionally added as a new API extension
>   * Adding a property to a resource representation
>   * Adding an optional property to a resource representation which may
>     be supplied by clients, assuming the API previously would ignore this property
> 
> This seems to say that you need an api extension to add a value to a
> returned dict but not to add a new header. So that would imply that
> checking the headers should allow additional properties but checking
> the body should not. Is that the desired behavior?

On " Generally Not Acceptable" of https://wiki.openstack.org/wiki/APIChangeGuidelines ,
the above case is not mentioned. So I'm not sure we have already gotten
a consensus that we should not allow additional properties of response bodies.
I has gotten this point twice through patch reivews. Just IMO, it is OK to
extend response bodies with additional properties. That means "additionalProperties"
should be True in JSONSchema definitions.
The removal/rename of properties cause backward incompatibility issues and
Tempest needs to block these changes, but additional properties don't cause
the issues and clients should be assumed to ignore them if not necessary.

This is just my opinion, and it is not strong one. So I also know some
feedbacks about this.


Thanks
Ken'ichi Ohmichi