Version Control: How Do Programmers Keep Track of Stuff?

Since developers are human, software isn’t a fire-once-and-done matter. To allow for these imperfections, software comes in “versions”.

Version Numbers

To keep things simple, programmers tend to release program versions with numbers on them, using a few possible conventions.

The first, and oldest, convention is to name versions by their relationship to a programmer’s finish line:

  • “Alpha” versions are barely-working programs that were quickly pushed out (e.g., v0.36). Features are sparse and it barely works.
  • “Beta” versions are nearly complete, but may still have bugs (e.g., v0.80).
  • “Final” versions are finished, but there are often little details that need tweaking afterward (e.g., v1.02).
  • After that, most developers tend to make version numbers based on features they’ve added. On a large project with many developers, it’s not uncommon to add letters and extended numbers afterward (e.g., v2.58.7216.R35).

Alternately, if you have a particularly long-term software development cycle, you can use the year and month the version is released instead (e.g., v.2019.10).

In general, there are 3 categories of update, which may have different versioning:

  1. Major updates, which will make the API integrations no longer work the same.
  2. Minor updates, which will allow the older versions to be “backwards-compatible”.
  3. “Patches”, which make backwards-compatible bug fixes.

One long-term solution is to roll out beta/testing versions with odd numbers (e.g., v1.25) and stable versions with even (e.g., v1.26).

In long-term software where there’s a certain level of public exposure for that software (such as entire operating systems), the marketing people will often make catchy names to reflect the version (e.g., Precise Pangolin, Big Sur, Lollipop).

Control Software

To keep track of changes while creating software, most programmers like to use a version control system such as Microsoft’s GitHub or Apache’s SubVersion.

Version control systems let software engineers figure out when code was updated, so they can trace where they may have made a mistake.

Further, many version control systems track each line of code, so they can easily trace exactly what happened.

Git

To manage versions the creator of Linux, Linus Torvalds, created Git. It accommodates versions held on various systems, and is portable enough to be used by any large organization.

The flow of activity in a normal IDE is pretty straightforward with Git:

  1. Modify code as needed, with the IDE of your choice tracking the changes.
  2. “Staging” the information to be ready for the final code. This way, people can review it to make sure it’s error-free. Make any more changes, and those can be staged as well.
  3. “Commit” the code, which makes it populate the changes to all the “repositories” of that code. The developer usually leaves notes about what changed as well.

In a strictly practical sense, a Git version control is simply a file folder, but with a hidden .git folder that keeps logs of all the changes. A developer can also have a .gitignore file with regular expression conditions that prevent certain files from being tracked by Git.

GitHub

While Git is a completely open-source experience, GitHub is a private organization that began as a startup to host Git instances. It scaled upward to become like almost any other large-scale company, then got bought out by Microsoft in 2018.

From that point, GitHub has become a part of Big Tech. Their most advantageous offering is GitHub Copilot, which scraped all the GitHub code to create an AI-assisted model to help with coding.

Forks

When someone is developing software, there are a variety of reasons to make an alternative version of that software:

  • Cleaning up code to make it more stable, but it’ll take a long time.
  • Someone else wants to add features of their own, or wants to take over the project.
  • The developer gave up on that implementation of the project and wants to start over, but other people are depending on it.
  • The developer lost his mind, and someone has to continue developing with a “known-good” implementation of the code.

For whatever reason, the developer can “fork” the version. This new version is identical at first, but is meant to change differently than its original version.

One of the most common ways to fork is to “checkout” a previous version. It will allow rolling back to a previous version without committing. That second version can be committed separately without affecting the master/main. When/if the developer wants to “merge” them together, they simply have to run a console command.

Phasing Out

Unfortunately, all the changes in software mean that it’s not usually sensible to support only the latest version:

  • Feature-rich things like programming languages often have many plugins, libraries, and optional features that many users may not want to update right away when the current version is “known-good”. This is especially true with enterprise software, where an update can be very expensive and time-consuming.
  • Often, users can fit into archetypes based on how they update. Some people want a stable experience and will be slower to update, while others want the latest immediately.

For that reason, a project cycle often will “end-of-life” certain versions once they know 100% that a newer version has everything the older version has, though sometimes the developer will have rebuilt the next version of the system to be closed-source.

Sometimes, a version has to be “yanked” if it failed miserably. Maybe that version could easily hacked, or it might have crashed core features. For whatever reason, certain version numbers become very dangerous to keep around without patching or uninstalling/reinstalling with another version ASAP.

Some enterprise software can be so feature-laden and complex that most managers won’t feel comfortable bringing it to version 1.x. This can cause extreme consternation in developers as the project persists, and can provoke rewriting the code from scratch or trying to close out all bug reports. This psychological effect can also express with releasing to 2.x, 3.x, and so on.

Further Reading