Re: [irsg] [rfc-i] Poll: RFCs with page numbers (pretty please) ?

[Carsten Bormann <[e-mail]>]

On 2020-10-27, at 12:33, Keith Drage <drageke=[e-mail]> wrote:
> 
> I do agree that referencing by section numbering is the appropriate method.

Yes.

> However, as you say, it gets difficult with overly long sections, and this is something on which no guidance or checking is currently given. Perhaps it should be?

“Overly long sections” are by definition of “overly” not a good idea.
The question when a section is “overly” long is the interesting one then — there certainly is no algorithm that is going to tell you the answer.

> Further there are difficulties with section references if you allow text between x.y.z and x.y.z.1. Does a reference to x.y.z mean the entirety of the text including x.y.z.1, or merely the text before x.y.z. Other organisations restrict such usage in order to make section references unambiguous.

»Please see the introduction of Section 2 in RFC 8610 for a definition of the terms “vector”, “record”, “table”, and “struct”.«

Works for me (except that I can’t put in the right quotes).

> I also note people are referring to the XML version as the definitive version, but surely the section numbers do not appear until the presentation version is created?

There is no “the XML version”.  There is the author-side XML (often created with a tool like kramdown-rfc).  This is edited, then run through a “prep-tool” which does all the formatting that is not rendition-specific, such as assigning section numbers (as well as reference anchors etc.), putting in boilerplate etc.  The result is the “canonical” XML, which is certainly no longer editable.  Then the rendition-specific formatters generate HTML, TXT, PDF from that.

Many of the problems we have with RFCXMLv3 stem from confusing the editable author-side XML form with the canonical, non-editable publishing form.  Both are expressed in XML, and they do use much of the same vocabulary, but they are *not* the same formats.

Grüße, Carsten