feat: RFC and sub-series aware ref sorting

[kesara]

Fixes #1318

[cabo]

Hi Kesara, please do not solve this with a ton of brittle special cases. Please do read up on: [Natural sort order](https://en.wikipedia.org/wiki/Natural_sort_order) [Sorting for Humans : Natural Sort Order](https://blog.codinghorror.com/sorting-for-humans-natural-sort-order/) You can simply use Python's natsort to use a debugged implementation. [natsort · PyPI](https://pypi.org/project/natsort/) Grüße, Carsten

[kesara]

@cabo, Thanks for the suggestion I wasn't aware about it.
I've created an alternate PR #1321

But there are issues with using natsort library in xml2rfc.
We are adding another third party dependency to xml2rfc.
At the moment it's active but the project seems to have a single maintainer.
I guess it's all good as long as we accept the risks involved with that project.

[cabo]

Hi Kesara, glad to be of service!

On Apr 14, 2026, at 01:55, Kesara Rathnayake ***@***.***> wrote: We are adding another third party dependency to xml2rfc.

Right. The essence can be written in a few lines of code. I don’t speak Python, but a friendly bot said: ``` import re _num_re = re.compile(r'(\d+)') def natural_key(s: str): """ E.g. "file000abc" -> ["file", (0, -3), "abc"] """ parts = _num_re.split(s) key = [] for i, part in enumerate(parts): if i % 2 == 1: # digit run value = int(part) key.append((value, -len(part))) else: # text key.append(part) return key def human_sort(strings): return sorted(strings, key=natural_key) ``` I prefer actually sorting leading zeros as text, so my regexp would be [1-9][0-9]* But of course in this context the sorting of RFC0020 vs. RFC20 may argue for going with the bot’s solution. Grüße, Carsten

[kesara]

Closing this in favour of #1321