[rfc-i] RFC citations committee I-D issued

Brian E Carpenter brian.e.carpenter at gmail.com
Fri Feb 11 13:17:19 PST 2011

On 2011-02-12 08:35, Joe Touch wrote:
> On 2/11/2011 11:16 AM, Scott O. Bradner wrote:
>> There is a basic question that I've not seen asked in this discussion:
>> what is the reason that someone wants to cite an ID in an RFC?
> FWIW, that is the basis of many of my comments.
> I would claim that most citations are one or both of:
> a) attribution
> b) supplemental content
> IDs ought to be limited to (a) only, 

I really don't see how you can make that assertion. The IETF can
and does prohibit any form of technical dependency on I-Ds
in IETF standards track documents. But the IETF has nothing to
say about non-IETF documents of any kind, and has never in practice
prevented informational citation of I-Ds for any reason - attribution,
content on which the specification does not depend, background information,
path not followed, etc.

I'd be all in favour of rewriting the relevant part of RFC 2026 to
formally establish the distinction between normative and informative
references in IETF standards track documents and to make it clear
that I-Ds must never be normative references. That's what we do in
practice anyway.

But I really don't see the point in trying to subdivide informative
citations of I-Ds into various categories, of which only some are
allowed in IETF stream RFCs.

In any case, the question the citations committee was asked was not that.
It was, given that an I-D is to be cited (non-normatively), for any
reason and in any IETF stream, or elsewhere, what should the
citation look like?

in the same spirit as e-mail
> citations. The citing doc should include a sufficient summary - either
> inline, or in an appendix - to stand alone in the absence of obtaining
> the ID.

That's a matter of opinion. However, if somebody needs to cite an I-D
in a paper that is size-limited by the publishing venue, they won't
necessarily be able to do that. This doesn't concern RFCs but it
does concern many conference and journal papers.

> You raise a different issue:
> 1) citing a document instance
> 2) citing a document group
> I.e., citing a single draft, or citing the chain of versions for which
> any version is likely sufficient. I would claim that this doesn't make
> sense, because you can't know that a future version might not omit the
> important info you want to cite (even for attribution).
> IMO you can't cite a group to indicate the "most recent" version; that
> is nonsensical as a reference. References should be persistent and
> specific.

That depends on context, but in general I agree, which is why we ended
up with the specific date and the full filename as useful attributes
in a citation.


More information about the rfc-interest mailing list