Open Stack

On 06/10/2013 11:26 AM, Dolph Mathews wrote:
>
> On Mon, Jun 10, 2013 at 9:23 AM, Sean Dague <sean at dague.net
> <mailto:sean at dague.net>> wrote:
>
>     On 06/10/2013 09:48 AM, Alvaro Lopez wrote:
>
>         Hi all.
>
>         I've submitted a change [1] for Keystone whose tests are
>         failing, since
>         some tempest changes are made asumptions bases on the old behaviour.
>
>         The proposed change increases the length of the user's name in
>         the DB
>         up to 128 characters, instead of 64. Tempest is trying to create
>         a user
>         with a username of lenght 65 and expects it to fail.
>
>         How should I proceed? Should I submit a change to tempest to fix
>         this,
>         even before knowing if my original change seems reasonable to be
>         merged?
>         Should I wait for my change to get some feedback before sending the
>         modified test for tempest?
>
>         Regards,
>         Alvaro.
>
>         [1] https://review.openstack.org/#__/c/22694/
>         <https://review.openstack.org/#/c/22694/>
>
>
>     Typically the route is propose the tempest change, and get core
>     contributors from the core project (this time keystone) to +1 it
>     (saying that they agree with the change).
>
>     The core team for keystone would need to believe that this doesn't
>     represent an API change without a version bump. And that
>     python-keystoneclient would do the right things with a new client on
>     an older system. But I'll leave that up to them.
>
>
> This definitely doesn't represent an API change, and would be backwards
> compatible with existing clients (this validation is only performed
> server side, anyway).
>
> Following Chmouel's comment on the review, I'm wondering if it would be
> possible to preserve the existing default but make this value
> configurable moving forward (which would mean no changes in tempest).
> I'll save that discussion for the code review, but wanted to mention the
> possibility of not affecting tempest here.

Making this configurable seems like a bad idea from a cross cloud 
compatibility perspective. If your application that creates users 
depends on a certain user name size, then can't be used on other clouds 
because they set their default too low, then that would be an issue.

Why not just bump it to 255, which is a reasonable db threshold, and be 
done with it.

	-Sean

-- 
Sean Dague
http://dague.net