To date (or at least in the past 1.5 years I've been working on
Hadoop), our shortest timeline from feature freeze to a release vote
has been around 4 weeks. This time is typically spent fixing blockers.
I'd like to propose that we tighten up on what a blocker is, get the
release out sooner, and then plan to do follow-on bug fix releases on
a regular interval. As your release manager, I'll sign up to
consider a bug fix release every 2 weeks if there is a critical
enough bug or a critical mass of bugs. Roughly every 2 weeks, then,
would be a window to consider a bug fix release.
For 0.16.0, can we set a goal of having all blocker committed and
calling a vote next Thursday, Jan 24, with a blocker defined as
a) a bad regression
b) an unintended incompatibility
c) bad bugs in an important new feature
Thoughts?
Nige
PS As a reminder, 0.17.0 is scheduled for the first Friday in April
(3 months).
PPS Monday, Jan 21 is a holiday in the US for many companies
