The existing signature matches

https://docs.python.org/3/library/xml.etree.elementtree.html#xml.etree.ElementTree.Element

I think there is value in keeping that.

I can choose to keep the old signature if you prefer, but what do you think about the call sites? (essentially, is element(..., attrib=...) really better than element(..., **...))?

I think attrib= is more readable. The advantage of kwargs is that you can specify them directly, e.g. …, id=…, but that doesn’t work for keys containing non-identifier chars. So we are left with {“some-key”: … }. Normally, passing a dict through via ** is ok, but since the API of ElementTree provides the attribute arg and we have the choice, I’d stick with that.

Note: I assume that’s the intention of the ElementTree design: use attrib for more complex configuration and direct keywords as a convenience for very simple cases like (…, id=…).

Do you want to just close this, or are there still parts that you think are worth keeping? I'm fine either way.

Making XMLWriter private is still reasonable. Choose yourself, whether closing+separate PR or force-pushing is simpler.

Kept only the "Make XMLWriter private" part, plus converting a few attributes into direct kwargs (the hopefully uncontroversial ones, with no dict unpacking).

Doc failure appears related:

/home/circleci/project/lib/matplotlib/backends/backend_svg.py:docstring of matplotlib.backends.backend_svg.XMLWriter:1: WARNING: py:class reference target not found: matplotlib.backends.backend_svg._XMLWriter [ref.class]

Hopefully fixed now.

Sadly not, it seems.