Bastien <b...@altern.org> writes: > Hi Suvayu, > > thanks for sharing this suggestion and to make it so clear. > > I understand the model you describe and I see why it's appropriate for > projects like "git" -- as IIUC, your proposal is very close to the one > described by git's maintainer. >
> - The latest git HEAD is stable enough so that many people live on it, > and can send feedback on patches. > It's all based on the idea "if it works, don't break it". > > But I'm open to any change if (and when) we need it. Here's my take on Suvayu's proposal. Personally I don't think it makes sense to keep separate topic branches in the public git repo for major parts of org-mode functionality since that would require merging the branches to master to get all of the functionality. Topic branches (even for the git project) are generally short-lived and for the developer's current set of patches until the feature is merged into master. At that point the developer normally deletes the topic branch. Junio (the git maintainer) keeps topic branches in his git development repo for active topics which are merged to maint, master, next, and pu integration branches. Only the integration branches are normally pushed to the public git repo that everyone clones. When a topic branch is merged to maint, or master it is normally deleted. Lots of people use org-mode for different aspects of the functionality it provides - I doubt anyone uses *all* of the features available in org-mode - I know I don't. The main advantage I see of keeping the current model is that people (including me) run directly from the tip of the master branch - I usually update at least weekly. This has the advantage that users who do this are testing the current development codebase and reporting problems early as they are encountered. This helps to fix problems early before it's time to create a new release. With Suvayu's model you would only merge for the release and that means the new master won't have the same level of testing exposure before a release is created. We don't have full coverage of org-mode's features in the ERT testing framework (far from it) and the more day-to-day testing we get during development the better. Regards, -- Bernt