cabo

The v2v3 output sure did for 3.11.1; I can't easily make a controlled experiment here.

The work happens In ``` text = lxml.etree.tostring(self.root.getroottree(), encoding='unicode', doctype=doctype_string, pretty_print=True) ``` So this might really be an lxml fix.

This is really weird. I have instances up to 3.11.1 that have both numbered=true and toc=default. I also have 3.11.1 instances that are toc=default only. I cannot find instances beyond...

I think you could also ask the RFC-editor to delete the spurious redundant attributes.

> I did; they declined. Whoa. Is this documented somewhere (outside an AUTH48 exchange)? Did they give a reason?

Maybe they are doing this for some internal tool. -*- mode: grep; default-directory: "~/std/rfc/authors/" -*- Grep started at Sun Feb 27 01:28:40 The below was a bit shocking for me......

See also I don't expect quick progress -- there are some 135 tickets, and three are scheduled now... So let's see what we can do on the authoring side.

> @mnot You should be able to emulate previous behaviour to reduce diffs by using `kramdown-rfc2629 -2` option. Thanks for finding the reason... Well, the workaround to go to -2...

Makes a lot of sense, probably alongside the existing options. Would need: -- a way to either translate existing naming or a commitment from the authors to specref-only naming, --...

I can reproduce neither issue with the current version of kramdown-rfc. draftr appears to be using an older version (I can't quite make out which), and for some reason it...