[Gluster-Maintainers] Release scheduling and lifecycle of versions

[Niels de Vos]

After several weeks of discussions on this list, in the weekly meeting
and ad-hoc IRC chats, I have come to the following understanding of the
release schedule, lifecyle and minor updates.

## What we want to address

1. regular releases, predictable cycle for users and other projects
2. faster "go to market" with new features, receive feedback from users
3. stable version(s) for 'happily running, no risk' deployments
4. active releases can do monthly bugfix updates (see backport criteria)


## Results of several rounds of discussion

It seems that a 3 month release cycle is the most attractive. Each
alternating major release would be a Long-Term-Support (LTS) one, the
others are short living versions for users that are eager to test out a
new feature.


## Three stable/LTS versions, 1.5 years of bugfixes

In dates and versions, this results in something like:

 3.5: EOL when 3.8 goes GA (2016-06)
 3.6: EOL when 3.9 goes GA (2016-09)
 3.7: EOL when 3.10 goes GA (2016-12)
 3.8 [LTS]: EOL when 3.14 goes GA (2017-12)
 3.9: EOL when 3.10 goes GA (2016-12)
 3.10 [LTS]: EOL when 3.16 goes GA (2018-06)
 3.11: EOL when 3.12 goes GA (2017-06)
 3.12 [LTS]: EOL when 3.18 goes GA (2018-12)
 3.13: EOL when 3.13 goes GA
 3.14 [LTS]: EOL when 3.20 goes GA

 (one of the 3.x releases will be called 4.0, I do not want to predict
  when that is ready and leave that up to the 4.0 team)

And, 'visualized':

 ----.
 3.5 |
 -----------.
        3.6 |
 ------------------.
               3.7 |
 ----------------------------------------------.
     |                   3.8                   |
     '------.------.---------------------------'
            | 3.9  |
     :      '------+-----------------------------------------.
                   |                   3.10                  |
     :      :      '------.------.---------------------------'
                          | 3.11 |
     :      :      :      '------+-------------------------------------
                                 |                   3.12
     :      :      :      :      '------.------.-----------------------
                                        | 3.13 |
     :      :      :      :      :      '------+-----------------------
                                               |                   3.14
     :      :      :      :      :      :      '------.------.---------
                                                      | 3.15 |
     :      :      :      :      :      :      :      '------+---------
                                                             |     3.16
     :      :      :      :      :      :      :      :      '---------

  2016-06   :   2016-12   :   2017-06   :   2017-12   :   2018-06

         2016-09       2017-03       2017-09       2018-03


## Two stable/LTS versions, 1 year of bugfixes

Looking at the diagram makes me wonder if this really is a sane approach
that we can promise to our users. When 3.13 (or 4.x) gets released, we
would have 4 active versions. At the moment we are already struggling
with three active versions, so I doubt we can really do that. So, here a
variation, and my suggestion to keep two stable releases instead of
three:

 3.5: EOL when 3.8 goes GA (2016-06)
 3.6: EOL when 3.9 goes GA (2016-09)
 3.7: EOL when 3.10 goes GA (2016-12)
 3.8 [LTS]: EOL when 3.12 goes GA (2017-06)
 3.9: EOL when 3.10 goes GA (2016-12)
 3.10 [LTS]: EOL when 3.14 goes GA (2017-12)
 3.11: EOL when 3.12 goes GA (2017-06)
 3.12 [LTS]: EOL when 3.16 goes GA (2018-06)
 3.13: EOL when 3.13 goes GA
 3.14 [LTS]: EOL when 3.18 goes GA

 (one of the 3.x releases will be called 4.0, I do not want to predict
  when that is ready and leave that up to the 4.0 team)

 ----.
 3.5 |
 -----------.
        3.6 |
 ------------------.
               3.7 |
 --------------------------------.
     |            3.8            |
     '------.------.-------------'
            | 3.9  |
     :      '------+---------------------------.
                   |           3.10            |
     :      :      '------.------.-------------'
                          | 3.11 |
     :      :      :      '------+---------------------------.
                                 |           3.12            |
     :      :      :      :      '------.------.-------------'
                                        | 3.13 |
     :      :      :      :      :      '------+-----------------------
                                               |            3.14
     :      :      :      :      :      :      '------.------.---------
                                                      | 3.15 |
     :      :      :      :      :      :      :      '------+---------
                                                             |     3.16
     :      :      :      :      :      :      :      :      '---------

  2016-06   :   2016-12   :   2017-06   :   2017-12   :   2018-06

         2016-09       2017-03       2017-09       2018-03


Providing bugfixes for a LTS release for a year seems like a suitable
middle road. If other projects/organizations want to support a certain
version longer, they can do so themselves or step up and contribute a
maintainer to our community.


I expect that the number of patches in the stable releases get reduced A
LOT compared to the current 3.7 version. With regular releases there
should not be a need to backport complex patches or features.
Maintainance of a stable release should be a very boring and simple
task.

Any suggestions for modifications or clarification needed before this
can be transferred to a good looking webpage (by Amye?) on gluster.org?

Thanks,
Niels
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://www.gluster.org/pipermail/maintainers/attachments/20160610/614f6416/attachment-0001.sig>