Open Stack

Hi Jay,

Thanks for your reply.

2017-03-29 12:26 GMT-07:00 Jay Pipes <jaypipes at gmail.com>:
> On 03/29/2017 02:04 PM, Ken'ichi Ohmichi wrote:
>>
>> Hi
>>
>> I have some questions about plancement API design from
>> https://review.openstack.org/#/c/376200
>> Current implementation of the above patch (PS26) adds GET method for
>> checking trait existence, and tags.rst of api-wg guideline[1] requires
>> GET for checking trait existence.
>> On the other hand, http.rst of api-wg guideline[2] requires HEAD instead.
>>
>> So there are two questions.
>> 1. Why the part of tags.rst is different from http.rst?
>>     I checked the review history of
>> https://review.openstack.org/#/c/155620/ which have added the part,
>> but I could not find the reason.
>> 2. trait should follow tags.rst? or http.rst?
>
>
> This is from the http.rst guideline in the API-WG:
>
> "TODO: HEAD is weird in a bunch of our wsgi frameworks and you don't have
> access to it. Figure out if there is anything useful there."
>
> and frankly, I tend to agree with the above statement. I'd say use GET for
> exists checking.

Ed submitted a LP about this as
https://bugs.launchpad.net/openstack-api-wg/+bug/1677360
On the LP, Chris did put a nice comment like:

  "HEAD is a shortcut to use when you don't want to pay the price of GET"

We can use HEAD without receiving a response body to check a resource
existence if the response body could be huge on GET.
In this case, IIUC trait API doesn't return a response body when
getting the detail of resource, that would be the same as a resource
existence check.
Both GET and HEAD could be acceptable in this case.
I was confused when seeing the conflict between tags.rst and http.rst,
I will try it clear later on api-wg guideline.

Thanks