Answering #99435 (comment):
To me the 5 first commits are the most important as they add new information.
The first commit is a code style issue in the tests; it does not add new information; it does not alter the semantics of the tests or code in question. I agree it is a bit of a code smell, but it is not worth fixing just because it's code smell. If (or when) we touch similar parts of the code, for example because we are adding new tests, or because we are adjusting incorrect tests, we can sneak in such a style change. (Also, this change has nothing to do with the issue title.)
These sentences are part of the reference for cursor methods. The reference for cursor methods describe what happens when you invoke these methods. I'm not sure explicitly stating that for every method is worth it, and I believe it clutters the docs.
I'm not sure this change is needed either. The docs say a transaction is always open; this implies a connection is open when the connection is created. Otherwise that sentence would have been incorrect.
Also, the current wording starts off simply by stating what behaviour the user can expect: A transaction is always open. Then it goes on adding details about the two most important methods, commit and rollback. The rewording in d864934 starts off by explaining the details and ends with explaining the behaviour. I'm not sure that is a good rephrasing. Also, the commit title does not reflect the actual change: it does not add connect to a list of transaction-opening methods; it does a complete rewording of a paragraph in addition to mentioning connect().
commit() and rollback() is already recommended in legacy commit mode; why is this needed? Rewording this again seems like churn to me.
IMO, this is not at all clarifying. PEP-249 does not describe an autocommit (non-transactional) mode. All it says is: some databases has some kind of concept of autocommit. If your underlying database supports such functionality, you can enable it when autocommit is True. PEP-249 leaves autocommit to the underlying database, and so do we. I'm strongly -1 to this change.
That is what I did by splitting this PR into 9 commits.
Unfortunately, that does not help; discussing and reviewing 9 separate issues in a single PR/issue is not productive. It is difficult to discuss (cross-conversations), and it is impossible to review, hence my closing the PR and the issue. If you find issues, create instead targeted and narrow issues, so we can discuss each change isolated.
Now for the rest of the commits:
Which inconsistency is this? I cannot deduce this inconsistency from the commit title "improve wording consistency". Looking at the commit diff, it makes many inconsistent changes; some stylish are rewordings and some are incorrect plural/singular adjustments. I don't see any explanation or comment regarding what's wrong with the current wording.
Ditto
Such a change might be worth considering, however see Guido's recent response to a similar concern.
IMO, this change is stylish churn.
I do appreciate you being interested in improving the docs, but please try to identify and explain the problems with the current docs, preferrably one by one, instead of proposing multiple unrelated changes like this.