Open Stack

On 06/03/2015 08:07 PM, Andreas Jaeger wrote:
> On 06/03/2015 03:57 PM, James Page wrote:
>> [...]
>> After some discussion with Thomas on IRC, I think this is more than
>> one effort; The skills and motivation for developers reviewing
>> proposed packaging changes needs to be aligned IMO - so I think it
>> makes sense to split the packaging teams between:
>>
>>   Debian/Ubuntu + derivatives
>>   CentOS/Fedora/RHEL + derivatives
> 
> So, would this be two packaging teams - on DEB team and one RPM team?

Yes.

> How would those two teams collaborate - or is no collaboration needed?

I don't see what there would be collaboration on. Even the thing we work
on is called differently (ie: "specs" in Red Hat, and "source packages"
in Debian/Ubuntu). Though if there are things on which we can converge
(like package names maybe?), it would be a good thing for final users.

> How are you handling Debian and Ubuntu differences?

There's not so much difference already on the Python module
dependencies, as Ubuntu imports packages from Debian. They imported 100%
of my work for Juno, then diverged for Kilo since I didn't upload (as
Jessie was frozen). Now that's the only difference we need to "fix", and
James is already working on that.

We already work on some packages together (like James Page and myself
are co-maintaining rabbitmq-server (he did most of it, shame on me)).

As for the core OpenStack packages, what I wrote works already on Ubuntu
if you rebuild the package, though it may need a few fix here and there
for version dependencies (as Jessie was frozen after Trusty), but that's
easy to fix. The harder part will be for Neutron and Nova. In case we
don't agree at all, we can still use "dpkg-vendors --derive-from ubuntu"
to check on which distribution we are, and do actions from there.
There's already such mechanism in the debian/rules file of the Debian
Nova package for example (to check on the group name for libvirt which
differs in Debian & Ubuntu). It's not in the Ubuntu version of the
package, but I'm sure that it will soon land there! :)

The only thing we can't fix will be the Maintainer: and Uploaders:
field, but even that, maybe we can find a way so we don't have to "fix"
it either.

So in a smaller amount of words: I'm sure we can find a way to reduce
our differences to the absolute zero on the source package level, even
though the result may look different thanks to some conditionals where
it can't be avoided.

Cheers,

Thomas Goirand (zigo)