[rfc-i] Informational and Experimental RFCs can have Normative references

Barry Leiba barryleiba at computer.org
Wed Apr 2 08:39:07 PDT 2014

> The *requirement* to separate the references is not actually enshrined
> in any IETF BCP, as far as I know.

Sort of yes and no.  The letter of what you say is correct: the
requirement to separate the references into different sections is in
an IESG statement:

But the logical separation of references into normative ones and
informative ones is in RFC 3967, at least (see Section 1.1), and that
logical separation is what drives the physical one.

The first two paragraphs of 3967 Section 1.1 are particularly
important to this discussion, so here, for easy reference:

   Within an RFC, references to other documents fall into two general
   categories: "normative" and "informative".  Broadly speaking, a
   normative reference specifies a document that must be read to fully
   understand or implement the subject matter in the new RFC, or whose
   contents are effectively part of the new RFC, as its omission would
   leave the new RFC incompletely specified.  An informative reference
   is not normative; rather, it provides only additional background

   An exact and precise definition of what is (and is not) a normative
   reference has proven challenging in practice, as the details and
   implications can be subtle.  Moreover, whether a reference needs to
   be normative can depend on the context in which a particular RFC is
   being published in the first place.  For example, in the context of
   an IETF Standard, it is important that all dependent pieces be
   clearly specified and available in an archival form so that there is
   no disagreement over what constitutes a standard.  This is not always
   the case for other documents.

Note in particular that "a normative reference specifies a document
that must be read to fully understand or implement the subject
matter," and that whether a reference is normative "can depend on the
context" of the RFC, perhaps being different for standards than for
"other documents."

To me, that says that Informational documents (and Experimental, but I
more often see the argument about Informational) definitely *can* have
normative references in them, and those normative references are the
ones that "must be read to fully understand [...] the subject matter."

> With my Gen-ART and Independent Stream reviewer hats on, I tend to
> be very critical of excessive use of Normative References in
> Info/Exp documents, because I think it gives them a spurious air
> of authority. But that's just me. (It's also vanishingly rare for
> an Info/Exp document to be promoted to standards track without
> being updated.)

I really don't follow that.  Authority doesn't come from the
references (and, really, I could argue that *being RFCs* gives them a
spurious air of authority, and that we should have a different series
for Informational documents, but that's a separate discussion).  But
it's critical that people reading Informational documents know which
references they need to chase for essential information (the normative
ones), and which "[provide] only additional background information."

> At the RFC Editor level it seems to me it should be something that
> is allowed but the precise policy is per-stream.

As I think should be the case for almost all policies relating to
document contents, beyond obvious formatting.


More information about the rfc-interest mailing list