First, thanks for your answer.
On 11/4/20 7:31 AM, Herve Beraud wrote:
> Greetings,
>
> Is it time for us to move some parts of Oslo to the independent release
> model [1]?
>
> I think we could consider Oslo world is mostly stable enough to answer
> yes to the previous question.
>
> However, the goal of this email is to trigger the debat and see which
> deliverables could be transitioned to the independent model.
>
> Do we need to expect that major changes will happen within Oslo and who
> could warrant to continue to follow cycle-with-intermediary model [2]?
I would hesitate to try to predict the future on what will see major
changes and what won't. I would prefer to look at this more from the
perspective of what Oslo libraries are tied to the OpenStack version.
For example, I don't think oslo.messaging should be moved to
independent. It's important that RPC has a sane version to version
upgrade path, and the easiest way to ensure that is to keep it on the
regular cycle release schedule. The same goes for other libraries too:
oslo.versionedobjects, oslo.policy, oslo.service, oslo.db, possibly also
things like oslo.config and oslo.context (I suspect contexts need to be
release-specific, but maybe someone from a consuming project can weigh
in). Even oslo.serialization could have upgrade impacts if it is being
used to serialize internal data in a service.
Agreed, the goal here isn't to try to move everything to the independent model but more to identify which projects could be eligible for this switch.
I strongly agree that the previous list of projects that you quote should stay binded to openstack cycles and should continue to rely on stable branches.
These kinds of projects and also openstack's services are strongly tied to backends, their version, and available APIs and so to openstack's series, so they must remain linked to them.
That said, many of the others can probably be moved. oslo.i18n and
oslo.upgradecheck are both pretty stable at this point and not really
tied to a specific release. As long as we're responsible with any future
changes to them it should be pretty safe to make them independent.
Agreed.
This does raise a question though: Are the benefits of going independent
with the release model sufficient to justify splitting the release
models of the Oslo projects? I assume the motivation is to avoid having
to do as many backports of bug fixes, but if we're mostly doing this
with low-volume, stable projects does it gain us that much?
Yes, you're right, it could help us to reduce our needed maintenance and so our Oslo's activity in general.
Indeed, about 35 projects are hosted by Oslo and concerning the active maintainers the trend isn't on the rise.
So reducing the number of stable branches to maintain could benefit us, and it could be done by moving projects to an independent model.
I guess I don't have a strong opinion one way or another on this yet,
and would defer to our release liaisons if they want to go one way or
other other. Hopefully this provides some things to think about though.
Yes you provided interesting observations, thanks.
It could be interesting to get feedback from other cores.
-Ben