[rfc-i] Is there a use case for 2119 keyword markup?

Paul Hoffman paul.hoffman at vpnc.org
Wed Jun 18 16:46:32 PDT 2014

On Jun 18, 2014, at 4:50 PM, Heather Flanagan (RFC Series Editor) <rse at rfc-editor.org> wrote:

> A question came up recently regarding whether there were any serious use
> cases around semantically marking up RFC 2119 keywords (when used _as_
> keywords) in the new format. 
> In the HTML draft, it says:
> 3.3.1.  Requirement Keywords
>   The RFC2119 keywords in the document will be set off with special
>   markup.  They are surrounded with a <span> element containing the CSS
>   class rfc2119.  For example:
>   They <span class='rfc2119'>MUST</span> be surrounded
> For this to happen, we need to add something to the XML vocabulary as
> well.  Does anyone have a use case where this kind of markup would be
> useful, or is it just a "nice to have, because we can, but not if it
> increases the overall cost of creating RFCs"?

As a note, Heather has asked this a few times in the design team, and I haven't seen a use case other than "it will be easier for people to see when reading the RFC". That feels untrue to me, given that we have all gotten quite used to pattern-matching on capitalization.

> Note: Whether or not we decide to add markup around the keywords, the
> current guidance around capitalization, etc, as described in RFC 2119
> will still apply.

If we have two completely ways to indicate "this is a 2119 word", then there is a very real possibility that there will be confusion that can lead to lack of interoperability. If a draft author turns in a draft with all of the 2119isms in caps, and half of them marked, but half not, what is a reader of that spec supposed to assume about the unmarked ones? Worse, what if the RFC Editor muffs one of the 2119isms in an RFC by capitalizing but not marking it?

Without a strong semantic use case, this seems like a bad idea. And I say that as one of the few people who sometimes makes a living pulling out the MUSTs and SHOULDs from RFCs for conformance and interoperability testing.

--Paul Hoffman

More information about the rfc-interest mailing list