Open Stack

On 02/22/2016 07:00 AM, Sean Dague wrote:
> On 02/21/2016 01:41 PM, Jay Pipes wrote:
>> On 02/21/2016 12:50 PM, Chris Dent wrote:
>>>
>>> In a recent api-wg meeting I set forth the idea that it is both a
>>> bad idea to add lots of different headers and to add headers which
>>> have meaning in the name of the header (rather than just the value).
>>> This proved to a bit confusing, so I was asked to write it up. I
>>> did:
>>>
>>>       https://review.openstack.org/#/c/280381/
>>>
>>> When I did, the best example for how _not_ to do things is the way in
>>> which we are currently doing microversion headers.
>>>
>>> So two questions:
>>>
>>> * Is my position on header non proliferation right?
>>
>> Yes, I believe so.

+1

>>
>>> * Is it so right that we should consider doing microversions
>>>     differently?
>>
>> Ship has sailed on a number of things, including this. I *do* think it
>> would be great to just use OpenStack-API-Version: $SERVICE_TYPE X.Y,
>> however we'll need to add another microversion to support that of
>> course. Isn't it ironic? Don't you think?

i'm wondering if the ship hasn't sailed on this as well. i also think it 
might be a little thrashing for us to oscillate back and forth on these 
headers. otoh, they aren't *widely* implemented yet, still i'm not even 
sure we could put this genie back in the bottle.

i think it might be nice to discuss this one more time at the next meeting.

>
> Actually, the headers can't be fully fixed in a microversion, because
> they are deep in the negotiation. We're stuck maintaining the old
> headers pretty much forever.

i not following the "deep in the negotation" part, would you mind 
expanding on this idea a little more?


regards,
mike