Registry requirements should not use RFC2119 language

[marcoscaceres]

Sections that are informative shouldn't be using normative language, unless we make the language forceful, but is can't be MUST/SHOULD/MAY etc.

[timcappalli]

Where does it say that the registry section is informative/non-normative?

[hober]

Where does it say that the registry section is informative/non-normative?

Registry sections are required to be non-normative by the W3C Process Document; see §6.5.2. Publishing Registries. Quoting that section:

A registry report or registry section is purely documentational, is not subject to the W3C Patent Policy, and must not contain any requirements on implementations. For the purposes of the Patent Policy [PATENT-POLICY], any registry section in a Recommendation track document is not a normative portion of that specification.

[npdoty]

That text is a little confusing to me. For the purposes of the Patent Policy it's not normative, but registries are required to include registry definitions, and the definitions must set out requirements for constraints and policy on adding values to the registry, which sounds like it should include normative language.

I think the "must not contain any requirements on implementations" means implementations of the specification, not entities that are potential values that might be listed in a registry. I think we should use normative language to detail the requirements (on the Working Group, on proposers, etc.) for what ends up in the registry, but that doesn't put any requirements on any piece of software that implements the surrounding API specification.