Protect Part of the Code

Question

My company works with different developers from different studios, always sharing the code to everyone aboard. At the moment, we need to protect a part of the code, not from theft but from some random developer in the studio maybe changing something important by mistake. Is there a way to protect a critical part of the code from this happening?

Answer 1

You need some mechanism that prevents developers from changing parts of the code they shouldn't be changing, at least not without review. This means that you will have to integrate your version control server with some access control or code review system.

Git in particular has no built-in access control model and a fairly weak security model (it is easy to fake a commit that looks like it was created by a colleague). That means you need an external system.

Such a system could be implemented using server-side hooks that for example check whether the commits are authentic (e.g. you could enforce that commits must be signed) and the hooks could reject a push if code outside of that developer's authorization was changed. Such hooks could also enforce that the code was first reviewed by a different person.

GitHub in particular already implements various mechanisms to that effect.

• Branches can be marked as protected, so that no one (or only admins) can push directly to that branch. All others must first create a pull request for their changes, and someone with write access to the GitHub repository must approve the pull request before it will be merged.

• You can create a CODEOWNERS file which associates file patterns with groups of reviewers. These people will be automatically alerted when a pull request touches the code they own. This helps to keep code reviews useful for larger repositories.

  See also the announcement on the GitHub blog. The system was inspired by the review system used for the Chromium project.

A completely different yet similar approach is the Lieutenant model used by the Linux Kernel. Everyone can change anything in their own repositories, as is normal with Git. But these changes will not generally become part of the official release. Instead, developers have to ask a Lieutenant to include their changes in their repository, until the changes bubble up to the official repository. At each level, reviewers sign off on the commits to ensure they are OK. The effect is similar to the CODEOWNERS approach, though less centralized and more reliant on personal trust.

Git as a distributed version control system is more flexible and less restrictive by design than centralized systems. More enterprisey centralized systems such as Perforce do feature a detailed access control model, for example preventing access to some directories and branches, or only allowing read access. Such approaches are especially valuable when developers aren't trusted, for example when you need to hire consultants for a tiny module of your Top Secret World Domination Project.

Answer 2

I just want to suggest a simple solution for specific case.

always sharing the code to everyone aboard is meaning you are using Source Control like TFS for me. If it is correct, I can suggest an easy way to you with one assumption:

If different developers from different studios are responsible for separate module(s)(class library), but all modules work with together, then you can authorize each studio for their own module(s) on Source Control and publish all modules latest build version on a server. By doing this, each studio can reach their own source code from Source Control and can add required DLLs from that server. So, no one can change another studio codes.

Answer 3

As others have already mentioned, Git(hub) doesn't really have a deep access control mechanism, so you can't just lock files based on user access rights (afaik). Other VCSes might do this, but companies tend not to want to change their VCS willy nilly.

With Git, you'd be stuck doing manual code review in order to detect changes in files you don't want changed. Hopwever, I don't consider this a good enough solution, as it still leaves the door open to human error.

---

If you don't want a developer to touch the source code, don't give them the option to update the source code. You could, for example, provide them with the DLL and documentation. That documentation could even be in the form of a copy of the source code (with no way of committing changes, just a simple file copy with no Git access).

Whenever my company has subcontracted work (part of a project, rather than a whole project), we have exposed our source material using an (on-our-premise) Nuget repository so the subcontractors have access to, but cannot change, the actual source material.
You don't need Nuget for this. Any method of transferring a DLL to the studio would do (email, FTP, ...). But Nuget (among others) makes it easier to handle versioning in case you update your source code (e.g. while the studio is mid-development).

Depending on the studio, they can either have their own Git repository which they check in and you have accesss to; or they could decide to not give you access to their source code either, and in that case they simply return their own dll to you. The difference here depends on the contract between your company and the subcontractor, and who owns the studio's source code that is created for this project.

In either case, you can then run the end product using your own DLL. Even if the studio somehow altered the DLL you provided them, you can make sure that their illegal changes are undone by simply using your own version of your own DLL.