Re: draft-congdon-radius-8021x

[Thomas Narten <narten@us.ibm.com>]

> I think that the biggest difference is that "SHOULD" needs to be 
> represented in the interoperability report for progression to Draft 
> Standard, and "should" does not.

Not clear to me at all, though I have heard some folk make that
argument.

Consider also that there are PS documents that do not use 2119
language at all. What do they do for interoperability testing?

Also, I think its pretty silly to look at a spec and just do the
SHOULDs. I.e., folk sometimes ask me if they can just look at the
MUSTs (or SHOULDs) and only implement those, as a way of minimizing
what they have to do but still be compliant. My response is that that
would be seriously nuts and would almost certainly result in
implementations that don't work/interoperate. Document authors just
don't use SHOULD language consistently enough to rely on them for
drawing these sorts of conclusions.

Somewhat side note. When the IAB annulled the IESG decision to publish
IPv6 Addr Arch at draft, at least some (one for sure!) of the IAB
members indicated that lack of 2119 language made the document
insufficiently clear (which was the general argument that was made for
annulling the decision). I don't personally agree with the view that
2119 language is required, and I don't beleive the community would
either. And it was not the IAB consenus view (from what I can
tell). But I mention it in that at least some people in the community
assume that 2119 language is necessary for clarity.

Thomas