[dns-operations] [Ext] Questions on DNS Flag day 2020 proposal

Paul Hoffman paul.hoffman at icann.org
Mon Jul 8 13:47:55 UTC 2019

On Jul 8, 2019, at 6:32 AM, Tony Finch <dot at dotat.at> wrote:
> I thought SHOULD was for cases where there might be reasons outside the
> spec for not doing the thing, whereas I am talking about how to
> distinguish between SHOULD and MUST when the specification is not
> explicitly clear about the difference.

Nonono. Well, at least not for many of us. When used properly, SHOULD is used when the exceptions to the rule are given right there in the text. In good modern specs, it is always "SHOULD X except when Y", even though the "except" clause might come later in the paragraph or even later in the section.

As for the meaning of pre-2119 MUST and SHOULD, this was heavily discussed as 2119 was being formed. There was a brightly-blazing lack of consensus about it. Some important pre-2119 RFCs had their own definitions of MUST and SHOULD, but most didn't. (The face-to-face meetings that led to 2119 did have some farcical comedy to them; I remember someone saying something like "when we say 'should' in Swedish, it means...".)

--Paul Hoffman

More information about the dns-operations mailing list