Adopt the “time window-based” policy for support of Python versions from NEP 29

[mkoeppe]

The NEP 29 policy (https://numpy.org/neps/nep-0029-deprecation_policy.html) aims to "standardiz[e] a recommendation for project support of minimum Python and NumPy versions [to] improve downstream project planning."
With this ticket, we adopt this deprecation policy of old Python versions for the Sage library.

According to NEP 29 support for Python 3.7 should be dropped in December 2021. Actually doing this will be subject of a new ticket that should be merged at the beginning of the 9.6 release cycle.

Related discussions:

• https://groups.google.com/d/msg/sage-devel/fgsSsJmVVXo/n2BQBSaBAAAJ (2020-05)
• https://groups.google.com/d/msg/sage-devel/aggikzrEfUg/_fIz4W-0BgAJ (2020-08)
• https://groups.google.com/g/sage-devel/c/j1cwbTU8aOU (2023-02)
• on PR Adopt the “time window-based” policy for support of Python versions from Spec-0 #35403 (2023-03–)
• on PR Drop support for Python 3.8 #35404 (2023-03–)
• Vote: https://groups.google.com/g/sage-devel/c/3Zoq0CNE1hE (2023-05–26–2023-06-07)

CC: @slel @jhpalmieri @orlitzky

Component: python3

Author: Tobias Diez

Branch/Commit: public/python3/nep29 @ edf1711

Issue created by migration from https://trac.sagemath.org/ticket/30384

[mkoeppe]

Description changed:

--- 
+++ 
@@ -3,3 +3,6 @@
 In this case we should indicate our intention to follow this policy  in our documentation.
 
 
+Related discussions:
+- https://groups.google.com/d/msg/sage-devel/fgsSsJmVVXo/n2BQBSaBAAAJ
+- https://groups.google.com/d/msg/sage-devel/aggikzrEfUg/_fIz4W-0BgAJ

[mkoeppe]

Description changed:

--- 
+++ 
@@ -1,4 +1,4 @@
-Getting the Python 3.8 support in #27754 merged in 9.2 will allow us to adopt the NEP 29 policy (https://numpy.org/neps/nep-0029-deprecation_policy.html)
+Getting the Python 3.8 support in #27754 merged in 9.2 will allow us to adopt the NEP 29 policy (https://numpy.org/neps/nep-0029-deprecation_policy.html) for the Sage library.
 
 In this case we should indicate our intention to follow this policy  in our documentation.
 

[orlitzky]

comment:3

I'm happy to punt on this. +1 from me.

[tobiasdiez]

comment:5

What is needed to follow NEP 29? I voting on the sage dev mailing list?

[mkoeppe]

comment:6

https://wiki.sagemath.org/ReleaseTours/sage-9.2#Support_for_Python_3.6.2C_3.8.2C_and_3.9_added
already declares the Sage 9.2 conforms to NEP 29.

I think the only thing missing is to add some sentences to the developer manual.

I don't think we need a vote. In general there seems to be limited interest in discussing release policies.

[tobiasdiez]

Branch: public/python3/nep29

[tobiasdiez]

comment:7

Ok, I've added a few sentences. Thus, this should be ready for review now.

---

New commits:

964dd29

Add documentation for following NEP 29

[tobiasdiez]

Commit: 964dd29

[mkoeppe]

comment:8

I don't think that NEP 29 requires us to drop support for certain Python versions; rather, it allows us to do so after the listed dates and still be compliant.

Dropping support for a Python version would need a discussion on sage-devel, I would think. (We probably miscommunicated about this.)

[tobiasdiez]

comment:10

NEP 29 is actually quite explicit about that one should drop support after the 42 months, e.g "On Jun 23, 2020 drop support for Python 3.6 (initially released on Dec 23, 2016)". But that's, of course, only a recommendation (like everything in this NEP). It could be sage's policy to support also older python versions, but that's then no longer NEP 29 (in it's strict form). However, one should also keep in mind that supporting old Python versions results in technical debt that might be bigger than the gains from supporting old systems.

Also correct me if I'm wrong, but since sage is also installing its own python version, sage can still be used on older systems, right? It is just that the system python is not used.

Should I initiate a discussion on the mailing list, or do you want to do this?

[mkoeppe]

comment:11

I wouldn't like to drop Python 3.6 support in the Sage 9.3 release unless there are more substantial reasons, so I think it's better to have this discussion in the 9.4 cycle