[openstack-dev] [oslo] upgrade implications of lots of content in paste.ini

Doug Hellmann doug at doughellmann.com
Thu Feb 18 19:33:26 UTC 2016


Excerpts from Sean Dague's message of 2016-02-18 12:55:02 -0500:
> On 02/18/2016 12:22 PM, Michael Krotscheck wrote:
> > On Thu, Feb 18, 2016 at 9:07 AM Doug Hellmann <doug at doughellmann.com
> > <mailto:doug at doughellmann.com>> wrote:
> > 
> > 
> >     If the deployer is only ever supposed to set the value to the default,
> >     why do we let them change it at all? Why isn't this just something the
> >     app sets?
> > 
> > 
> > There was a specific request from the ironic team to not have headers be
> > prescribed. If, for instance, ironic is deployed with an auth plugin
> > that is not keystone, different allowed headers would be required.
> 
> Here is the future we're going to have.
> 
> Whatever the middleware does with no operator intervention will be how
> the world will work, and how you will need to assume the world will work
> going forward.
> 
> Right now, it appears that the default in the middleware is do nothing.
> That means CORS won't be in a functional state on services by default.
> However, I thought the point of the effort was that all the APIs in the
> wild would be CORS enabled.
> 
> I'm not hugely sympathetic to defaulting to not having the Keystone
> headers specified in the non keystone case. I get there are non keystone
> cases, but keystone is defcore. Making the keystone case worse for the
> non keystone case seems like fundamentally the wrong tradeoff.
> 
>     -Sean
> 

I agree. We should make this thing work for our needs first, and allow
flexibility on top of that. But the default should be made useful.

Doug



More information about the OpenStack-dev mailing list