[rfc-i] Categories of references
"Martin J. Dürst"
duerst at it.aoyama.ac.jp
Fri Oct 3 01:06:14 PDT 2014
On 2014/10/03 06:17, Brian E Carpenter wrote:
>> So how about a tripartition of
>> — normative (MUST read to implement this specification)
>> — informative (SHOULD read to implement this specification)
>> — historical (MAY read to understand historical background; NOT RECOMMENDED to read for implementers)
> Overkill. We have too much bureaucracy already. The text you supplied
> makes the status of such a draft 100% clear.
I fully agree with Brian and Heather. "MUST read" for normative is
usually quite wrong. If e.g. a spec uses HTTP, nobody will go and read
the whole HTTP spec. They'll just use a library. If every implementer
read every normatively referenced RFC, they'd never get anything
actually implemented. And if the text around the reference doesn't make
it clear whether and to what extend and for what purposes references
(normative or informative) should be read, then either the text may
benefit from improvement or the reader should get a reality check (or both).
> Historically, the only reason for splitting the references was to
> semi-automate the detection and avoidance of downrefs to I-Ds and/or
> non-standards. With good metadata we could have avoided even that.
> Unfortunately it's become a sacred cow.
> rfc-interest mailing list
> rfc-interest at rfc-editor.org
More information about the rfc-interest