Git and workflow models

Release of distributed version control systems made possible certain changes to workflow in software development. Some changes were immediately obvious while others might not have been quite so obvious. Here are some of them.

Essentially creating branches did not require long planning before as it used to be and could be created for even minor features or things that would be discarded later. Improved merging was the other part that made branches more useful: earlier method involved individual patch/diff files but version control had no ability to track what was already merged. When version control could track these for you it became smaller effort and allowed further changes.

Better merging allowed keeping different repositories: per sub-project, per developer, per location.. You name it. Each repository could have branches and changes independently of main project (“superproject”) which in turn would help to control user access (no need to commit to the main project), efficiency (different locations) and safety in case disaster strikes (additional clone). Having support for multiple repositories also allowed rewriting workflow rules entirely as suitable for the project, not as version control designer had set them.

Since every working copy is also a complete repository single developer could do all their work entirely locally, never needing to use another computer and you would have all you need on your laptop while travelling. Backups could be made from that one location but that would not be fully using the capabilities.

As with centralized systems it is entirely possible to also use Git with a central server: this is often used in corporate teams where the central model might be preferred due to way work is done and organized (timetables, resources..). But this is not a limitation as there can still be a repository for prepared releases separately from repository used during development or even multiple repositories for sub-projects. Essentially this is similar to old workflow with embellishments and integrates with backups in much the same way.

Real distributed workflows comes from the fact that no single repository is more “special” than any other: developers can pull from each other directly and/or push to another location for use by others. Because there is no longer limitation of a single repository as in the centralized model the ways of organizing workflow are much more open. In principle development can use any combination of branches, sub-projects and repositories for parts or whole of the development as needed for the workflow. In distributed model each clone of the repository is also a backup and each developer having a local clone also has a backup of the whole repository. Centralized server is not required, but can be still used for things like publishing regular builds of the project.

Largest impact of distributed versioning might still be ahead as people learn to use them and organize work in different ways. Distributed versioning systems have other interesting properties (merging separated from commit, tracking changesets of related changes instead of per-file changes) which could be discussed in length as well. In terms of managing software projects largest impact might be with how repositories are viewed: they are no longer as immutable as they were.

Related posts:

For quite a while in Canada, influencer advertising had the impression of being something too good to be true. Indeed, it is not true anymore! Influencer Marketing isn’t equivalent to it was 10 years…

2017 has traveled every which way, shaking up the digital money world with a swathe of newcomers sprinkled with stellar highs and some calming value revisions. Bitcoin finished the year having…

Songkick will automatically publish listings to the world’s most important music services, for free. We’ll push your tour dates to the Spotify concert’s tab and the ‘On Tour’ module with geo-located…