As a general rule, I agree with Gennady that it is better to have long-lived branches on the main repository than to have them in a fork, especially if the intention is for that branch to eventually become part of the master release cycle. With the long-lived branch in the main repository, we get to have the record of the git history and maybe more importantly there is transparency for the community on the full scope of activity that is happening on the project.
So I think the issue comes down to the nature of PRs. PRs to master have an increasingly higher standard that must be met prior to merge. PRs to a long-lived branch reside in (virtually) the same queue as PRs into master, but have an inconsistent and lower standard. So for repository maintainers that are studying the backlog of PRs should have a set of guidelines that they can apply.
If it is commonly agreed that having long-lived branches is good for transparency and coordination of different teams, then the reverse thinking can also be applied - are there features that are being developed either in forks (by committers) or in a long-lived branch where branch PRs are not currently being used? It's been Codenvy's approach for developers to just direct-commit to that branch until the entire branch is readied for the merge into master, and then we end up with a titanic singular PR.
I am indifferent to what the policy should be. I just think that it's time for the maintainers to agree on a simple set of rules, and then ask all of the committers to follow the same guidelines.
So I think:
1. More long-lived branches are a good thing
2. Using PRs to manage changes into a long-lived branhc is ideal when >2 people are contributing
3. We should document the labeling and management rules for branch PRs
4. We should document the standard of achievement for merge into a branch PR, even if that standard is "no minimum code quality test required"
5. We should place these rules into the Che wiki
We have the weekly Che planning call Tuesday morning Pacific - I would like this to be the discussion topic for part of the session.
Tyler