Thanks for providing lots of materials here. We will take a look, and possibly re-open the previous research effort on relevant topics here and there.

Currently, the file locking is the reason we use portalocker, because we do not want to get ourselves into the file locking business. We will need some investigation to see what we can do here.

Has there been any update on this? Upstream portalocker is now at 3.0.0 which conflicts with the version constraint in setup.py. Since most Linux distributions will always update to the latest upstream version of the packages they maintain, this breaks the msal-extensions package.

Has there been any update on this? Upstream portalocker is now at 3.0.0 which conflicts with the version constraint in setup.py. Since most Linux distributions will always update to the latest upstream version of the packages they maintain, this breaks the msal-extensions package.

The PR #117 was a ready-to-test 2 years ago, but it seems there was no strong interest. Now I rebased the PR and further polished it. @glaubitz or everyone interested can briefly test it by:

1. Start from a clean virtual environment and then pip install https://github.com/AzureAD/microsoft-authentication-extensions-for-python/archive/refs/heads/optional-portalocker.zip. Run pip list to confirm that portalocker is not intalled.

2. Run this one-liner to further confirm MSAL-Extensions is using the new file-based lock.

   python -c "from msal_extensions import CrossPlatLock as L; print(L)"
   <class 'msal_extensions.filelock.CrossPlatLock'>
   

3. You can actually use this branch with your MSAL-powered app and see whether it still works normally without portalocker.