Re: [Ietf-and-github] [Ext] GIT WG disposition

[Carsten Bormann <[e-mail]>]

On 2021-06-24, at 20:23, Spencer Dawkins at IETF <[e-mail]> wrote:
> 
> A day or two before that, I tripped over a bug that kramdown-rfc2629 currently produces v2 rfcxml (aka xml2rfc v2 input),

(That is not the bug.)
Kramdown-rfc2629 actually produces output that is optimized for conversion by xml2rfc into whatever xml2rfc happens to consider to be the v3 of the day.
(Using the --v3 flag on kramdown-rfc2629 enables using all of the v3 features and gives up v2 compatibility for that.)

> but xml2rfc command-line by default assumes --v3. If it sees v2 input, it internally runs the v2->v3 converter,

Exactly, but that is designed to eat v3-in-a-v2-shell as produced by kramdown-rfc.

> so the tools team needs to change the datatracker's submit code (which uses xml2rfc as a library) to run v2v3 when it sees v2 now instead of just running the v2 parser.

Right, *that* is the bug (i.e., trying to stick to pre-transition software when the <rfc> element doesn’t happen to look just right).
(Note that the pre-transition software is *not* being used by the RFC production center, so this approach is not only the wrong one to run a transition, but also produces late surprises when publishing the document as an RFC.)

> This one was discussed in a TRAC ticket, by Robert Sparks' request (at https://trac.ietf.org/trac/ietfdb/ticket/3305).

Indeed, and I look forward to getting that fixed.

I still recommend generating the .v2v3.xml before going to datatracker's submission interface, so there is more control over the generation process; it has not always been clear what version of xml2rfc is running under the submission process (sometimes weird environmental issues such as locale mess this up).

(If the above sounds a bit jaded:  The transition strategy used by xml2rfc works exceedingly well, and I have a lot of respect for Henrik and others getting this right at such an amazing level of precision.  It’s just incredibly frustrating when that great tool then is wedged into a submission process that doesn’t handle the transition as well.)

Grüße, Carsten