On 2019-10-24 14:32:03 -0500 (-0500), Ghanshyam Mann wrote: [...]
- Projects can start dropping the py2.7 support. Common lib and
testing tools need to wait until milestone-2.
[...]
This doesn't match the intent behind what I originally suggested nor my subsequent interpretation of what we discussed, and unfortunately the plan on the etherpad is slightly vague here too. I thought what we were agreeing to was that leaf projects (services and the like) had *until* milestone 1 (~2019-12-12) to remove Python 2.7 testing if they depend on shared libraries which are planning to remove support for it in Ussuri. From then until milestone 2 (~2020-02-13) shared libraries could work on dropping support for Python 2.7. If libs are allowed to drop support for it *after* milestone 2 then that doesn't leave much time before they're released at milestone 3 to stabilize or reverse course.
Phase-1: Dec 09 - Dec 13 R-22 Ussuri-1 milestone ** Project to start dropping the py2 support along with all the py2 CI jobs.
This is a milestone later than I expected, unless you mean they should be done by this point. It's just about removing jobs, so projects should be on the ball and do this quickly.
Phase-2: Feb 10 - Feb 14 R-13 Ussuri-2 milestone ** This includes Oslo, QA tools (or any other testing tools), common lib (os-brick), Client library. ** This will give enough time to projects to drop the py2 support.
This leaves less than 2 months where libraries are allowed to complete the necessary work before they get released for Ussuri (remember the final release for libraries is at R-6, the week before milestone 3).
Phase-3: Apr 06 - Apr 10 R-5 Ussuri-3 milestone ** Final audit on Phase-1 and Phase-2 plan and make sure everything is done without breaking anything. This is enough time to measure such break or anything extra to do before ussuri final release.
[...]
Libraries are released the week before this, so no, that doesn't really provide any auditing opportunity.
I apologize, in retrospect I realize that the distinction between "by" and "at" in English could be too subtle for a lot of folks to pick up on, and I should have been more explicit in my original proposal.