On Tue, Aug 1, 2017 at 1:23 PM, Ben Pfaff <[email protected]> wrote: > I'd like to create branch-2.8 today in preparation for releasing OVS 2.8 > later in August. I sent the patch that would lead off the branch a few > minutes ago: > https://mail.openvswitch.org/pipermail/ovs-dev/2017-August/336544.html > > According to our published schedule, we should have create branch-2.8 at > the beginning of July, but we delayed to allow some more features to > enter the 2.8 release. There continue to be reasonable arguments along > these lines in favor of delaying branching further: > > * Many DPDK patches have been posted but are still under review > or awaiting application. > > * Multiple series of patches leading up to NSH support have been > posted but are still iterating through reviews. > > The primary argument in favor of branching quickly is to keep to a > regular release cadence. My understanding is that multiple downstream > projects rely on prompt Open vSwitch releases. (This is mainly based on > speaking to Russell Bryant at Red Hat; he can fill in the details if > need be.) I'd prefer to stick to our schedule for this reason.
Yes, my reasoning is that a time based release schedule is only valuable if we stick to it as closely as possible. When dates are reliable, more downstream advance planning can be done. If we stray too far from published dates, downstream consumers will lose confidence in any published dates. The net result will be a longer window of time between an upstream OVS release and downstream integration. Here is a concrete example. We published a date for OVS 2.8 of mid-August. OpenStack Pike is scheduled to be released at the end of August. What version of OVS should a downstream aim to ship with this version of OpenStack? The ideal case in this scenario would be to provide OVS 2.8 with OpenStack Pike. Some new OVN features for OpenStack depend on support in OVN as well. The safer fallback is to plan on shipping OVS 2.7 instead. That fallback is much less sensitive to OVS release timing, but introduces another ~6 month delay before features are finally shipped to users via normal distribution mechanisms. A consistent time based release model can be very helpful to downstream consumers, but only if the dates can be relied upon. > > I propose that we branch today, with the following further details: > > * Continue to apply patches that add features, that as of now > have already been posted and ready for review or iterating on > reviews, until approximately this Friday. These patches will > need to be committed both to master and branch-2.8. > > * As always, bug fix patches are acceptable on any branch at any > time. > > * Plan to release, given acceptable quality, during the last > week of August. > > Thoughts? > > Thanks, > > Ben. > _______________________________________________ > dev mailing list > [email protected] > https://mail.openvswitch.org/mailman/listinfo/ovs-dev -- Russell Bryant _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
