Google Colab Python Version Update Breaks Code Compatibility

[kim-bongseok]

Describe the current behavior
The recent update to Python in Google Colab has made it incompatible with existing codebases that rely on specific Python versions or library versions. Downgrading the Python version using external scripts (such as py310.sh) partially resolves the issue, but doing so causes other critical dependencies or libraries to break, making the environment unstable and unsuitable for running many existing projects.

Describe the expected behavior
We expect Google Colab to maintain backward compatibility for users with existing projects, or to provide an easy and stable way to select Python versions without compromising the compatibility of other libraries or dependencies.

What web browser you are using
Google Chrome Version 132.0.6834.159

This issue has made it difficult for users to continue working on projects that rely on older Python versions, particularly when downgrading Python causes other critical library dependencies (e.g., numpy, numba) to fail.

We request a solution or a guideline to manage Python environments more effectively in Colab.

[metrizable]

Thanks for filing the feature request. It's an interesting idea and one we've considered.

During large upgrades of the runtime, we do provide the ability to select the fallback version from the Command Palette via the "Use fallback runtime version" option. In your case, it would allow you to select the prior runtime which would be Python 3.10. I appreciate that the fallback feature is slightly different than this request captures.

As an aside, in general, we try to keep the default runtime roughly in line with the latest Python version that's received its last bug fix release (in this case Python 3.11).

You may also be interested in the public images we publish on a roughly monthly cadence and maintain for six months (instructions for use) . Again, I appreciate that this feature is slightly different than your request, but it may fulfill a part of your needs.

[kim-bongseok]

Thanks, the "Use fallback runtime version" is indeed useful for now as it lets me run my code well. However, did I get it right that this solution would only work/be available temporarily?

[metrizable]

@kim-bongseok Correct. The fallback version for Python 3.10 is only available through mid-February. You'll need to work with the notebook authors to upgrade them to Python 3.11. The public images are available for six months.