On Wednesday, 31 May 2023 at 03:00:53 UTC+8 Nils Bruin wrote:

On Tuesday, 30 May 2023 at 11:13:27 UTC-7 tobia...@gmx.de wrote:

that we normally drop support for older versions right after this support 
window (i.e. also adapt the drop schedule 
https://numpy.org/neps/nep-0029-deprecation_policy.html#drop-schedule). 
I've formulated an improved formulation for this policy at 
https://github.com/sagemath/sage/pull/35403#issuecomment-1566110884 which 
should hopefully clarify this point.
 
(Dropping support means increasing `python_requires` and removing github 
tests for this old version, i.e. similar to 
https://github.com/sagemath/sage/pull/35404 which is modeled on the PR 
drooping Python 3.7.)


Formulated like this it would seem to me we could needlessly reduce the 
platforms that sage builds on: this suggests that the "drop Python 3.*" PR 
would get merged simply because it's time: in theory we could end up with a 
release where the sole commit is the dropping of support (in practice 
development is way too active for this). I would expect that such a PR 
would only get merged as a dependency of some other PR (that makes a fix, 
enhancement, or upgrades some other component).


Yes, in theory that could happen. Since its real work that someone has to 
put in to create a "drop Python 3.*" PR, I suspect that this persons needs 
to see a clear benefit before opening the PR. So I suspect that one first 
notices "oh shit, this doesn't work on Python x" and then proceeds to 
create a drop PR if its already past the NEP29 time window. But I'm happy 
to encode this practice in the policy. Do you have a suggestion for a 
formulation?
 


Having a policy that makes clear when such a drop PR is acceptable as a 
prerequisite for merging could save work and discussion. Numpy's schedule 
looks like a reasonable candidate for that.


That was exactly my intention. 

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-devel/d66e2ed7-b45c-430d-85e1-5789396b416en%40googlegroups.com.

Reply via email to