Re: [Rswg] draft-rswg-rfc7997bis-01: Do keywords and citation tags need to be in ASCII?

[John C Klensin <[e-mail]>]

--On Wednesday, January 24, 2024 04:55 +0000 "Salz, Rich"
<[e-mail]> wrote:

>> Using symrefs=no is the most striking demonstration of
>> contempt for the reader's time that an author can choose to
>> put into an RFC. I wish RPC style would no longer allow this.
> 
> I don't know if I'd go that far, but yeah, it sucks to see [3]
> and not remember if it's what you thought [1] or [4] were. It
> is disrespectful of the readers's time and cognitive load. And
> you really do expect the readers to greatly outnumber the
> authors

Rich, Carsten,

I can't match the vehemence of your comments without further
raising the temperature of the virtual room, but let me suggest
that there are several assumptions about writing style in your
comments and that it would be entirely reasonable for the RPC to
apply more stringent editorial criteria when numbers are used.

Is has become common practice in the IETF to write phrases
similar to

   ... as discussed in [RFC1149] and [RFC9999]...

that sends the author off on a quest to figure out what on earth
RFC 1149 and RFC 9999 are about.   There is no question in my
mind that 

   ... as discussed in [3] and [5]...

would be worse because the RFC numbers provide some clues.  On
the other hand, many of us, and probably a large proportion of
readers who are not active in the IETF, can't remember more than
a few (or even a few dozen) RFCs by number, so, even though the
first form clearly provides more of a clue than the second, how
much more may be questionable and depend on the reader.

That problem gets even worse --and the differences smaller--
when citing articles that are not RFCs.  Suppose there were an
author, Joe Shmoo, whose writings relevant to a specific RFC
were quite prolific.   Then the citations might end up looking
like
   [Shmoo2001]
   [Shmoo2002]
   [Shmoo2003a]
   [Shmoo2003b]
   [Shmoo2004]
and so on.  I suggest that no one other than Joe, and maybe not
him, is likely to remember enough about which papers where
written in which years to make
   
   ... as discussed in [Shmoo2003a]
 
at all more useful to the typical reader than

   ... as discussed in [7]
In Rich's terms, they both suck.

On the other hand, many style and grammar models dislike all of
those forms because citation (or footnote, or whatever) anchors
are not noun-type objects that can be used in that way.  In the
RFC Series, people who are even a bit fussy about that would be
(and have been) more likely to write:

	... as discussed in RFC 1149 [RFC1149] and RFC 9999
	  [RFC9999]...

if that form is used, the information loss from using 

	... as discussed in RFC 1149 [3] and and RFC 9999 [5]...

instead is minimal and probably very close to zero.  If we are
concerned about cognitive load or the user's time and remember
that RFC numbers do not, in general, provide strong semantic
clues, an even better sentence would be 

	... as discussed for avian carriers [RFC1149] and deeply
	submersible ones [RFC9999]

with even less information loss substituting the numbered
references.  At that point, the user may need to look at the
references for details, to check on sources or authority, and so
on, but does not need them to understand what the sentence is
trying to tell them.  One could (and probably should) go even
further with that style if, for example, there were two
well-known proposals for QoS with avian carriers and one was
considered, by the RFC author at least, to be, er, for the
birds.  Then one might have

	... as discussed in Waitzman's proposal for avian
	carriers [RFC2549]

So, without changes to the vocabulary and with the understanding
that it may be out of scope for RSWG, if we could encourage the
RPC (or they could spontaneously take it on themselves) to be
sure that, if an author uses numbered references, they use one
of the latter forms, with phrasing like "discussed in [3]..."
being completely forbidden.  I think that would put us in better
shape than forbidding numbered references and even than many of
the other forms above.   As you have probably guessed by now, I
would generally prefer the form shown above this paragraph,
whether it is used with bracketed RFC numbers or numeric
references, over "...as discussed in [RFC 2795]" for precisely
the cognitive and user time reasons you mention, but the that
form has become popular enough that I wouldn't dream of
prohibiting it after some flag day (well, I might dream, but...).

best,
   john