Discussion: change the release frequency?

[jmini]

Since the start of the project we have used following release schedule:

In average one release each week.

• Each month a release containing breaking changes with fallback:
  • 3.0.0 June 2018
  • 3.1.0 July 2018
  • 3.2.0 August 2018
  • 3.3.0 September 2018
• After each of those release, we provided 3 bugfix release. if we have a look at the 3.2.x release :
  • 3.2.0 2018-08-06
  • 3.2.1 2018-08-14
  • 3.2.2 2018-08-22
  • 3.2.3 2018-08-30

Is this release frequency appropriate ?

If the time between two releases was changed to 10 days or 2 weeks, would it impact your project ?

More generally, are you switching to a release when it is published ? Or are you waiting ? Are you using Snapshots builds ?

Feedback is appreciated.

We need to understand what we need to do about release cycles. Every input is welcomed.

[jmini]

I forgot to say that the consideration to change the release schedule is not for 3.3.0. A new schedule could be used starting with 3.3.1.

cc: @OpenAPITools/generator-core-team, @OpenAPITools/openapi-generator-collaborators

[wing328]

I vote for a bi-weekly patch release and bi-monthly minor release.

[grokify]

I like frequent releases as it helps get new fixes in and provides faster feedback for contributions. One or two weeks apart is fine as people as contributions will still be top of mind.

[wing328]

If we decide to update the release schedule, we'll need to update our documentation. Here is a list for tacking:

• https://github.com/OpenAPITools/openapi-generator/blob/master/docs/qna.md#why-was-it-decided-to-fork-swagger-codegen-and-to-maintain-a-community-driven-version
• https://github.com/OpenAPITools/openapi-generator/wiki/FAQ#how-long-does-it-take-to-release-a-stable-version-eg-305--306

[wing328]

OK. Let's start with the upcoming minor release 3.3.0, which has been rescheduled to Oct 1st (Mon) to follow the bi-monthly release schedule

[tobymurray]

Just chiming in as a user of this library - we're in the middle of migrating away from the Swagger generator. In our experience, the Swagger generator was difficult to upgrade and difficult to trust, so we've been upgrading slowly (or not at all).

The release frequency that would provide value is directly related to the project quality/stability. If there's a track record of patch releases not introducing regressions, I'd be sold on upgrading optimistically every time there's a new release. This is something that's easy to do with Guava or Apache Commons Lang or what have you. At present we haven't had enough experience with OpenAPITools to upgrade as aggressively, so twice-monthly is plenty fast enough for us (aligns with a 2 week sprint as well for our own planning).

[wing328]

If there's a track record of patch releases not introducing regressions, I'd be sold on upgrading optimistically every time there's a new release

Thanks for the feedback. So far we've not heard any report of regressions between patch release.

We hope the smaller patch release will be easier to upgrade.