You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
icalendar doesn't have a policy about support of unsupported versions of Python. I think it would be good to have one. I'd suggest that icalendar follow the cadence of Python. @niccokunzmann@SashankBhamidi@angatha@geier what do y'all think?
The criteria to close this issue include to update the following files, and potentially others:
I think there is a source file that conditionally checks type hint support for Python<=3.10, but I can't recall where. It might use logic with Python<3.11 instead.
Do we have a way to comunicate to contributors that they can use python 3.11 features without removing 3.10 from tox.ini, pyproject.toml, ...? I would keep it listed as supported as long as we realy do not support it. On the other hand, if we can use 3.11 features, we should do it.
@angatha I'm not sure I understand your comment. 7.x.x currently supports 3.10 - 3.14, inclusively.
For users that need 3.10 support, they may continue to use the 7.x.x line and earlier. They won't be able to get the features in 8.0.0. Because of that, I'd suggest we cut another 7.x.x release before an 8.0.0, so they get the benefit of the last few months of work that are itching to be released.
We deprecated Python 3.9 because there was a huge shift in the way we could do typing. It made development easier.
I do not see a need to deprecate Python 3.10, yet. Whoever finds it useful can put in the work.
Doing it for the sake of following Python deprecations limits usage. We have some download stats from PyPI for assessing that.
I think we should still add a Python support policy to the versions and branches documentation, to save having parts of this discussion again. Here's the start of that narrative. This would be the sole closing criteria of this issue. I'd appreciate any suggestions or feedback.
Python support policy
The stable version of icalendar supports and is tested on versions of Python that don't have a negative impact on its maintenance and security.
For example, icalendar 7.0.0 dropped support for Python 3.9 because it lacks adequate typing support that Python 3.10 provides.
icalendar may add support for stable released versions of Python.
Python itself has its own support policy.
icalendar may support versions of Python that Python itself no longer supports.
Although it doesn't affect icalendar's functionality, documentation currently depends on Python >= 3.11. Sphinx 9.1.0 dropped Python 3.11. To continue to get the features, especially accessibility, of Sphinx and the PyData Sphinx Theme, we'll need to bump up their versions and use at least Python 3.12. I'll make this comment a new issue for me to work on.
I don't see much advantages in the new features intriduced in 3.11. Therefore we may keep 3.10 supported. For 7.x.x, that is out of question. For 8.x.x I want the option to use 3.11 features, so 3.10 can't be officially supported. Otherwise, we can not just drop support. But I wan't to encurage developers to still support it.
Security policy states we only support the latest major version of icalendar. We allow backport of security fixes. Python version deprecation is a major version bump.
So the major version supported is the latest one. That happens as soon as we decide to deprecate a version. Generally, we do not deprecate officially supported Python versions.
There is a point that we do not promise to deprecate or add versions by our own energy but as is needed. However, we can decide to veto deprection of supported versions and always invite support of new Python versions. That is a sensible appoach.
Adding a new Python version is not a major bump.
Also, deprecating 3.10 is not a must for icalendar 8. I would like to go back to quick releases and just bump major versions when changes arrive.
On that note, I wonder if sqash merges make it easier to backport fixes.
Also, deprecating 3.10 is not a must for icalendar 8. I would like to go back to quick releases and just bump major versions when changes arrive.
You mean, there could be multiple major releases of icalendar? For example, we had the internal methods like encode that we made private and that would be enough to do a major release?
On that note, I wonder if sqash merges make it easier to backport fixes.
Sounds like a question about preferences to me. I reason this way: Backporting requires copying the changes/commits and adapting them. I see three realisit scenarios:
(no squash) cherry-pick all commits and add one commit to adapt the changes
(squash) cherry-pick the squash commit and add one commit to adapt the changes
(squash) cherry-pick the squash commit and then ammend the required changes
Option 3 may be confusing because the refernced commit in the commit message does not match the changes from that commit. So there are only options that add a commit. cherry-picking one or n commits does not make a difference, so none seems easier. Then it became just a question if you prefere squashing.
And about that (I do not use it):
Does github keep track of individual commits or will the commits get lost eventually? (if the branch is deleted for example).
Would it be an advantage for you, when for example the AI usage statement from one commit applies to all commits?
I think, in contrast to 3.9, 3.10 has been part of LTS releases, so we can expect that that hangs around for a while.
I have no preference - we can drop Python 3.10. I will create an alpha release first. Then, we can still add more breaking changes to that.
Activity
angatha commentedon Aug 8, 2026
Do we have a way to comunicate to contributors that they can use python 3.11 features without removing 3.10 from
tox.ini,pyproject.toml, ...? I would keep it listed as supported as long as we realy do not support it. On the other hand, if we can use 3.11 features, we should do it.stevepiercy commentedon Aug 8, 2026
@angatha I'm not sure I understand your comment. 7.x.x currently supports 3.10 - 3.14, inclusively.
For users that need 3.10 support, they may continue to use the 7.x.x line and earlier. They won't be able to get the features in 8.0.0. Because of that, I'd suggest we cut another 7.x.x release before an 8.0.0, so they get the benefit of the last few months of work that are itching to be released.
niccokunzmann commentedon Aug 8, 2026
We deprecated Python 3.9 because there was a huge shift in the way we could do typing. It made development easier.
I do not see a need to deprecate Python 3.10, yet. Whoever finds it useful can put in the work.
Doing it for the sake of following Python deprecations limits usage. We have some download stats from PyPI for assessing that.
stevepiercy commentedon Aug 8, 2026
@niccokunzmann thanks. I couldn't remember which version of Python held back typing.
Python 3.10 downloads of icalendar hover around 10%, whereas 3.9 is now less than 1%. These two screenshots illustrate trends over the last few months.
I think we should still add a Python support policy to the versions and branches documentation, to save having parts of this discussion again. Here's the start of that narrative. This would be the sole closing criteria of this issue. I'd appreciate any suggestions or feedback.
Python support policy
The stable version of icalendar supports and is tested on versions of Python that don't have a negative impact on its maintenance and security.
For example, icalendar 7.0.0 dropped support for Python 3.9 because it lacks adequate typing support that Python 3.10 provides.
icalendar may add support for stable released versions of Python.
Python itself has its own support policy.
icalendar may support versions of Python that Python itself no longer supports.
Although it doesn't affect icalendar's functionality, documentation currently depends on Python >= 3.11. Sphinx 9.1.0 dropped Python 3.11. To continue to get the features, especially accessibility, of Sphinx and the PyData Sphinx Theme, we'll need to bump up their versions and use at least Python 3.12. I'll make this comment a new issue for me to work on.
angatha commentedon Aug 9, 2026
What Nicco saiy 😅
I don't see much advantages in the new features intriduced in 3.11. Therefore we may keep 3.10 supported. For 7.x.x, that is out of question. For 8.x.x I want the option to use 3.11 features, so 3.10 can't be officially supported. Otherwise, we can not just drop support. But I wan't to encurage developers to still support it.
officially:
7.x.x: python 3.10 - 3.14
8.x.x: python 3.11 - 3.15?
inofficially / as long as easily possible:
7.x.x: python 3.10 - 3.14
8.x.x: python 3.10 - 3.15?
stevepiercy commentedon Aug 9, 2026
@angatha that's what I generalized in the draft Python support policy above. It would go after Branches.
I think official and unofficial support is confusing. I'd stick with a single support policy.
niccokunzmann commentedon Aug 9, 2026
Hm. These seem logically equal.
Security policy states we only support the latest major version of icalendar. We allow backport of security fixes. Python version deprecation is a major version bump.
So the major version supported is the latest one. That happens as soon as we decide to deprecate a version. Generally, we do not deprecate officially supported Python versions.
There is a point that we do not promise to deprecate or add versions by our own energy but as is needed. However, we can decide to veto deprection of supported versions and always invite support of new Python versions. That is a sensible appoach.
Adding a new Python version is not a major bump.
Also, deprecating 3.10 is not a must for icalendar 8. I would like to go back to quick releases and just bump major versions when changes arrive.
On that note, I wonder if sqash merges make it easier to backport fixes.
angatha commentedon Aug 9, 2026
You mean, there could be multiple major releases of icalendar? For example, we had the internal methods like
encodethat we made private and that would be enough to do a major release?Sounds like a question about preferences to me. I reason this way: Backporting requires copying the changes/commits and adapting them. I see three realisit scenarios:
Option 3 may be confusing because the refernced commit in the commit message does not match the changes from that commit. So there are only options that add a commit. cherry-picking one or n commits does not make a difference, so none seems easier. Then it became just a question if you prefere squashing.
And about that (I do not use it):
Does github keep track of individual commits or will the commits get lost eventually? (if the branch is deleted for example).
Would it be an advantage for you, when for example the AI usage statement from one commit applies to all commits?
niccokunzmann commentedon Oct 14, 2026
I think, in contrast to 3.9, 3.10 has been part of LTS releases, so we can expect that that hangs around for a while.
I have no preference - we can drop Python 3.10. I will create an alpha release first. Then, we can still add more breaking changes to that.
niccokunzmann commentedon Oct 26, 2026
Dateutil dropped Python 3.11. We can drop it, too.
If someone needs features backported, they can do it.
Also, this will propagate: recurring-ical-events and so forth.
@tobixen you depend on icalendar - how is caldav doing for 3.10? Required?
stevepiercy commentedon Nov 16, 2026
FYI, there's a backlog of things to handle with continued support of Python 3.10. I hit this one today:
-#2135 (review)
We have lots of checks for Python 3.10 in the code and other files now.