Which Workflow?

When considering which Git Workflow to use one is naturally drawn to three leading candidates:

1. Gitflow: because it is the most popular and everybody can’t be wrong - can they?
2. Oneflow: because it is said to address the shortcoming of Gitflow - well, that’s good
3. The workflow used by the Git Open Source Project (git.git for short); because surely the authors of Git know a thing or two about using it – don’t they?

Why not Gitflow?

There are several reasons why Gitflow is not going to work well for my circumstances:

1. The reality of Gitflow is it results in a dirty looking Commit History that is incredibly difficult to follow for anything more than a trivial project with very few released versions. This is argued by the author of Oneflow and by Contributors to Git itself
2. I work on Products that carry the legacy of Customer specific Variants and so consequently does not uphold the requirement that “every new production release is based on the previous one” as there are multiple, simultaneously released versions that are incompatible with each other

The first reason makes me disinterested in using Gitflow due to my limited capability to understand complexity (simpler is always better for me) and the second point just seals the deal as it effectively says that only a fool would proceed.

Why not Oneflow?

An obvious point would be if not Gitflow, then why not Oneflow. After all, Oneflow ought to overcome the issues related to a dirty history. That is its main reason for existence after all. The answer is that point 2 (see Gitflow above) still kills us with regards to Oneflow.

I would also have concerns that little regard is given within Oneflow for having many, parallel feature developments and extended periods required to stabilize my Products due to a set of changes being attempted. The git.git workflow does expend quite a lot of effort on this point; obviously it needs to because it is a big Open Source Project with many Contributors attempting many, many pushes.

Why not git.git?

There are several reasons why git.git won’t work for Maginus OMS:

• It is not an Open Source Project, so requiring an Integration Workflow utilizing the role Maintainers seems to add complication and a bottleneck
• Git.git does not detail well enough the interaction with topics/feature branches (as Oneflow does) such that Commit History remains in a clean and intelligible state
• Git.git does not address our needs in respect of Customer specific Variant releases

So which one?

I believe that what I need is a combination of the detailed Topic Branch policies of Oneflow (to keep the Commit History clean), the higher-level Branching Policy of git.git (to deal with parallel development, stabilization, release and hot fix related quandaries). Additionally, I need to give some thought as to how to sort out work related to Customer specific Variants.

I call this mash-up Maginus Flow. Not a very inventive name, but it’ll do for now.

---

Last modified on 2018-08-20