Open Stack

Hi,

We have been making progress on the scoring for the "TC direction"
columns of the capabilities[0], but we had two questions before we can
finalize:

A. Sometimes the capability name doesn't really reflect the backing
test. Should we judge the intent of the capability, based on its name
(and hope that the backing tests will be fixed later), or judge solely
on the backing tests (and hope that the capability name will be fixed
later) ?

Case in point: the "compute-auth" capability is backed (only) with the
"test_get_metadata_of_alt_account_server_fails" test (which makes sure
that if you try to get metadata for a server that isn't yours, that it
fails). That test has little to do with "auth"... the capability tested
therefore just seems to be a regression test on a security feature.

Our current proposed score is based on the test contents, so if the
capability was supposed to mean something else, please let us know.

B. The "Complete" criteria is a bit of a misnomer. My understanding is
based on Rob's post on [1], where "Complete" reflects that the test is
valid across all configurations and deployment of OpenStack. Is that the
correct interpretation ? Some TC members have been looking at it from a
"is the feature behind this test complete, or still being improved"
angle, and I would like to confirm my interpretation is correct. Links
to reference documents are welcome.

Thanks in advance,

[0] https://review.openstack.org/#/c/100722/
[1] http://robhirschfeld.com/2014/05/01/defcore-capabilities-selection/

-- 
Thierry Carrez (ttx)