Description
The current release schedule is ad hoc, which mostly follows a pattern that a bug gets fixed or feature gets added, and then several months later people get tired of depending on an unreleased feature and ask for a release. So I'd like to establish a predictable release schedule, while at the same time not artificially inducing churn or overburdening downstream projects with too short of deprecation timelines.
For the level of activity we see, I think quarterly minor releases would be sensible, and a yearly major release would allow us to begin to finally remove components that have been deprecated for years at this point. Major releases will have a minimum one month RC phase.
Bug-fix (micro) releases could continue on an ad hoc basis, on a monthly schedule if there's steady accumulation, shorter in the case of downstream breakage or longer for slow periods.
I'll plan on starting with a minor release (2.4.0) in March or April, and a major release (3.0.0) in September or October.
There's a related discussion to be had about Python 2 support, but perhaps that makes more sense to start another thread?
@nipy/core @nipy/team-nibabel Let me know if you have any reservations (or unreserved support...). I'll also ping @GaelVaroquaux, @pauldmccarthy and @afni-rickr to make sure nilearn, and the Pythonic sides of FMRIB and AFNI get a chance to weigh in. Please tag in anybody else that you think might want to comment.
@yarikoptic Any concerns from a NeuroDebian perspective?