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

Dave CROCKER dhc at dcrocker.net
Thu Feb 10 15:18:58 PST 2011

On 2/10/2011 2:23 PM, Joe Touch wrote:
> Hi, Brian,
> On 2/10/2011 2:16 PM, Brian E Carpenter wrote:
> ...
>>> I think MAY is OK (for the filename), but SHOULD isn't, for that reason.
>>> It gives a misimpression of publication; again, the point is to cite for
>>> credit only.
>> Not so, if you actually want to discuss the details of a versioned draft.
> IMO, it's incumbent on the author to include enough context for that discussion,
> not just to cite off to a doc that may be gone.

Part of that context is the citation name of the document.  In fact it is the 
single-most helpful part.

We need to be careful that a desire for some sort of philosophic purity does not 
get in the way of pragmatic utility.

The "draft-" string provides an unambiguous identifier.  Making it difficult or 
questionable to use that string in a citation is counter-productive, to the 
point of silliness.

This thread has pointed out that some variants of the name make sense, notably 
whether to include the version number.  The choice depends on the purpose of the 
citation.  (I assume that leaving the version off does not inherently mean 
"latest" but rather implies the whole set.  If we want to be sure that 'latest' 
is meant, we need a notation for it.)

A different issue is whether to include a URL.  Again, from a pragmatic 
standpoint, having a URL is far friendlier to the reader than not including it. 
  The problem is that the IETF does not provide a stable URL for I-Ds.  We 
should get that changed.

> Overall, it's dangerous to cite something that can disappear (except as credit)
> - esp. if it's intended to disappear.

I-Ds don't disappear.  They might move but they are still accessible. They 
actually never did disappear, but now it's convenient to get at the older ones.

> In this case, the date gets you the difference anyway. I'd say "MAY include the
> filename in general, SHOULD include the filename if there are multiple versions
> for a given date)

Joe, this isn't a abstract math exercise.  Making the reader play games to find 
the document is wrong-headed.  (Also, Julian just pointed out to me that the 
dates on these docs are not confirmed during submission...)



   Dave Crocker
   Brandenburg InternetWorking

More information about the rfc-interest mailing list