Re: [bfcpbis] Spencer Dawkins' Discuss on draft-ietf-bfcpbis-rfc4582bis-15: (with DISCUSS and COMMENT)

[Paul Kyzivat <[e-mail]>]

On 11/5/15 5:06 PM, Alissa Cooper wrote:

> "REQUIRED unless" strikes me as a faulty construction. That case is what
> RECOMMENDED is for, when there is a known exception to the requirement.

I don't think I agree.

I think "REQUIRED unless" is quite precise. Using "RECOMMENDED unless" 
says to me that in cases where the the "unless" doesn't apply then 
RECOMMENDED applies and hence can still be violated for cause.

Historically, SHOULD/RECOMMENDED has been used in a rather casual way, 
in the spirit of "we trust you to be responsible about this". And it 
hasn't worked out well. So now we at least *say* that you ought not use 
SHOULD/RECOMMENDED without being explicit about the conditions when it 
is acceptable to violate it. At that point, the distinction between MUST 
and SHOULD goes away.

I think the following constructions are most clear:

- You MUST do foobar except when blah blah blah.
- You are REQUIRED to do foobar except when blah blah blah.
- You SHOULD do foobar. The only acceptable situation to not do foobar
   is blah blah blah.
- It is RECOMMENDED that you do foobar. The only acceptable situation
   to not do foobar is blah blah blah.

I have had implementers tell me that when they get an assignment to 
implement a spec, their management tells them to only do the MUSTs, 
because that is all that is required to satisfy the checkoff!

	Thanks,
	Paul