Re: [dnsext] What is the 'canonical' spelling: RRset of RRSet ?

["Roy Arends" <roy@nominet.org.uk>]

Alfred Hönes wrote on 10/15/2009 04:01:37 PM:

> Folks,
> I need a bit of feedback for a review.
> 
> What is the correct/preferred spelling/capitalization from the PoV
> of the list:  "RRset"  or  "RRSet"
 ?
> 
> There's a very decided position in the IETF w.r.t. the capitalization
> of other important acronyms, like "IPsec".
> Do the readers of this list care similarly in the above case?
> 
> Background information usage in on selected 'important' DNS RFCs:
> o  "RRset" is used in RFCs 4033, 4034, 4035, and 4509;
> o  "RRSet" appears in RFCs 2181 and 5155.
> o  current DNS related I-Ds also vary; there are even drafts
using both.
> 
> Any opinions?

I can't believe this came up :-)

This is what I wrote a wrote many moons ago:

Begin forwarded message:

From: Roy Arends <roy@dnss.ec>
Date: February 16, 2007 9:37:46 PM GMT+01:00
To: DNS Directorate 
Subject: tongue in cheek weekend reading: RRSet vs
RRset 

Peter Koch and I are pondering which of the term RRSet
or RRset is the preferred one. In other words, should it be written with
or without capital 's'.

Have fun reading.


Dear Peter,

I did a little research on what the correct notation
should be.

RFC 1996 introduces the term RRset, but does not explicitly
define it.
RFC 2136 explicitly defines RRset.
RFC 2181 explicitly defines RRSet.

RFC 2181 did not re-define the term RRset to RRSet,
since it updates nor obsoletes RFC 2136.
We are left with two different acronyms that have
a similar definition, and we may assume that they can co-exist. However,
we still have to pick one for our current draft. To decide which is preferred,
we simply count the RFCs that use RRset and use RRSet, and we go with term
that is used in the most RFCs. Note that I did not include the obsoleted
drafts in my count.

RRset:
  1996 2136 2308 2663 2694 2874
  3007 3364 3403 3833 3958 4033
  4034 4035 4470 4471 4472 4509

RRSet:
  2181 3123 4592 4641 4697 4795

Thats 24 RFCs, of which

 18 RFCs use RRset,
  6 RFCs use RRSet.

This, however, is skewed. Since the 'RRset' RFCs might
actually refer to RFC 2181 (the 'RRSet' RFC), in which case those are completely
wrong and should be candidate for a -bis draft. With priority ofcourse.
 Since RFC 1996 and RFC 2136 were published before RFC 2181, I'll
skip those. These are in error: 2308 2874 4471

The adjusted count is now:

 15 RFCs use, or should have used the term RRset,
  9 RFCs use the term RRSet.

This is still skewed, since some RFCs refer to both
2136 and 2181. Since 2181 was published later than 2136, we assume it should
be leading. Those RFCs are ambiguous and should be candidate for a -bis
draft. With priority ofcourse. These are in error: 2694 4033 4034 4035
4472

The adjusted count is now:

 10 RFCs use, or should have used the term RRset,
 14 RFCs use RRSet.

But, this goes both ways. RFC 4592, which uses RRSet
(defined in 2181), does NOT refer to 2181, but DOES refer to 2136. This
obviously and blatantly wrong, and is thus candidate for a -bis document.
With Priority of course.

The adjusted count is now:

11 RFCs use, or should have used the term RRset,
13 RFCs use, or should have used the term RRSet.

There is a problem with RFC 2308 though. This RFC
was previously counted as wrong.  RFC 2308 has RFC 2181 in the reference
section, but does NOT refer to 2181 in the body. Since we now have RFC
2308 with an obsolete orphan reference, we must re-instate the status of
RFC 2308 as using the term RRset correct. The reference section needs to
be updated, hence 2308 is still candidate for a -bis. With priority ofcourse.

This now makes our count:

12 RFCs use, or should have used the term RRset,
12 RFCs use, or should have used the term RRSet.

Since there is an equal count of RFCs using (or should
have been using) the term RRset or RRSet, this approach does not lead to
a definite answer.
We are now back to the question: which term is the
prefered, 2136's RRset or 2181's RRSet.

Lets assume the choice is arbitrary, and we choose
RRset from 2136.

When our draft becomes an RFC, the count will then
be:

13 RFCs use, or should have used the term RRset,
12 RFCs use, or should have used the term RRSet.

Since most RFCs use the term RRset, I recommend that
we do a 2181-bis as well, after which we can -bis all RFCs that use the
term RRSet, with priority of course. This appears to be a lot less work
compared to revving all RFCs that use the term RRset in error.

[afterword]

There is one RFC that I haven't mentioned yet. RFC
2929 written by Don Eastlake and Bill Manning.

I have never EVER seen such an atrocity in my life,
"RR Set". An extra space. What were they thinking! Was the RFC
just too damn short ? Is life not confusing enough ?

Have a good weekend :)

Roy Arends
ignoramus savant.