RE: SHOULD vs MUST case sensitivity

In like mode, it would be nice if there was a series of guides that described how to go about specifying certain types of document such as we have for MIBs.

For example I would like there to be a document that described how to go about specifying a cryptographic algorithm for use in IETF protocols with instructions like:

* You must specify the cryptographic primitive it implements
* You MUST specify a canonical OID for identiyfing the algorithm
* You MUST specify a canonical URI for identifying the algorithm
* You MAY request assignment of protocol specific identifiers if the stated requirements for those protocols are met.
* You SHOULD specify test vectors
* You MAY include code that implements the algorithm

And so on.

________________________________

From: John Levine [mailto:johnl@iecc.com]
Sent: Tue 7/1/2008 11:40 AM
To: ietf@ietf.org
Cc: Hallam-Baker, Phillip
Subject: Re: SHOULD vs MUST case sensitivity

>* Whenever the keywords are used they are to be considered normative
>* Whenever the keywords are used they SHOULD be capitalized

Ahem:

* Whenever the keywords are used they MUST be capitalized

>* Editors SHOULD avoid use of normative keywords for non-normative
>language, even in drafts.

Yes, I agree.  I've done this in recent drafts that I've been working
on and found it to be quite effective.  Looking at each magic word and
thinking about whether it's telling the reader something important
about interoperating is good discipline, yet not terribly hard.

R's,
John