`Doc/c-api/init.rst` is way too big

[ZeroIntensity]

Documentation

While working on #141810, I noticed that init.rst covers three very complicated topics:

1. Interpreter initialization and finalization
2. Threads and concurrency
3. Subinterpreters

This makes it pretty huge and difficult to find things. I think we should split this up into two pages. Let's have a page for interpreter-related topics (initialization, finalization, subinterpreters), and a separate page for anything related to threads (PyMutex, critical sections, thread states, etc.).

Linked PRs

• gh-141811: Split up init.rst into multiple pages #144844
• [3.14] gh-141811: Split up init.rst into multiple pages (GH-144844) #145061

[StanFromIreland]

These moves are quite unfortunate for translations, but I’ll try to keep track of this and inform translation teams. Are you going to be backporting to 3.14?

[ZeroIntensity]

That would be ideal. Otherwise, any changes to the docs of those APIs will have conflicts when backporting. Is there anything that can be done to improve this for translations?

[vstinner]

I think we should split this up into two pages.

It could be 3 pages:

• Interpreter initialization and finalization
• Threads and concurrency
• Subinterpreters

[StanFromIreland]

Is there anything that can be done to improve this for translations?

Automatically, for non-Transifex translations (TX memory should merge them back in), no. It will require each translation team to manually merge the messages back in from an older commit. Due to this, I want to notify them ASAP after merge to recover the message and not re-translate things. I'll look into adding a guide for it to the devguide.

[ZeroIntensity]

It could be 3 pages

Sounds good. I'll work on this once #141810 is done.

[picnixz]

FTR, breaking the file might also break SEO and (possibly) intersphinx if we change the label names. See also #126053. One reason we still didn't break down some very large pages:

Try not to break all the external links going into the pages. We don't want to invalidate all the references from blogs, tweets, stackoverflow answers, etc.

[ZeroIntensity]

I think that isn't as important here; C API docs are going to be significantly less linked than something like datamodel.rst. But, according to that issue, we can redirect things anyway.

[picnixz]

Yeah but I would rather have the EB make a decision here instead of use core devs: see https://discuss.python.org/t/restructuring-the-stdlib-docs/102506.

[ZeroIntensity]

How do I ask the EB?