Open Stack

lzy.dev at gmail.com wrote:
> On Thu, Sep 12, 2013 at 9:43 PM, Thierry Carrez <thierry at openstack.org> wrote:
>> So, this is a significant feature... which paradoxically is a good
>> reason to accept it *and* to deny it. On one hand it would be nice to
>> complete this (with Glance support for it being landed), but on the
>> other it's not really a self-contained feature and I could see it have
>> bugs (or worse, create regressions).
> 
> Hello Thierry Carrez, two questions, whether we pass FFE or not.
> 1. why you think it's not a self-contained feature/patch, do you think
> the patch miss something?

By "self-contained" I mean something that almost doesn't touch existing
code, like a new backend driver. That means the odds of introducing a
regression in the existing featureset is very limited. Your patch
doesn't miss anything, it just touches enough existing code that the
odds of introducing a regression are not null.

> 2. I'd very like to know what's wrong in current patch # 33409, can
> you point the bugs out which you mentioned above?

My assessment above is not a code review. It's risk management. We've
been asking people to test and report bugs ever since we hit feature
freeze. If we change code paths, we basically throw away the existing
testing and increase the odds of introducing of regression. If we add a
feature late, it will see less testing and increase the odds of
introducing a bug.

The feature freeze is not about arbitrarily blocking people. It's about
maximizing our chances of producing a good release. Every time we accept
an exception, we increase the risk. The later we accept an exception,
the bigger the risk. The proposed feature may or may not be essential
enough to the success of the release to be worth that risk. It's a
trade-off, which is why we have an open discussion about it.

-- 
Thierry Carrez (ttx)